Find OpenCV Projects on Upwork with Vollna

We are seeking an experienced AI engineer and mentor to develop AI websites and personal AI ecosystems. The role involves creating AI-driven websites, integrating AI tools, and mentoring on AI adoption. The ideal candidate will have a strong background in AI engineering and a passion for mentoring. This is a part-time role with a project scale of small to medium, expected to last 1 to 3 months.

About the job
We are building a technical team to develop software solutions in the medical field. We are looking for specialists who want to work in a small team, making a direct impact on the product, with a strong focus on quality, data security, and scientific rigor.

OPEN POSITIONS:

1. Technical Lead

-Technical coordination of the team, system and data flow architecture, planning and tracking deliverables, quality assurance

-Over 5 years of experience in developing artificial intelligence / machine learning systems (target range: 5–10 years)

-Technical team leadership experience; Python, PyTorch, or TensorFlow

2. Machine Learning / Data Science Expert

-Dataset preparation, model development and training, statistical validation, and error analysis

-Minimum 2 years of experience in machine learning / data science

-Python (pandas, scikit-learn, PyTorch, or TensorFlow), solid knowledge of statistics

3. Computer Vision Expert

-Development and training of deep learning models on image data, quality control of image datasets, performance evaluation

-Minimum 2 years of experience in computer vision / deep learning

-Python, PyTorch or TensorFlow, OpenCV

4. Full-stack / Mobile Software Developer

-iOS and Android mobile applications, backend services and APIs, API integrations, data security and protection (GDPR)

-Minimum 2 years of experience in software development

-Native or Flutter / React Native; Python, Node.js, or Java; cloud services (AWS or Google Cloud)

Mandatory Requirements (All Positions)

-University degree in the field (Computer Science, Computer Engineering, Engineering, Mathematics-Informatics, Bioinformatics, or similar)

-Professional working proficiency in English

Nice to Have

-Master's or Ph.D. degree in the field

-Experience in R&D (Research & Development) projects

-Experience with regulated software (e.g., IEC 62304)

What We Offer

-Attractive compensation package, commensurate with experience

-24-month employment contract

-Full-time or part-time options compatible with other professional commitments

-100% remote work

-Direct impact on the product and a genuine research component

-Bachelor's degree (as listed)

We need senior engineers who can take an AI idea and turn it into something real that people use every day. You should be comfortable building with LLMs and data, and just as comfortable shipping the product around it. You'll work on AI agents, ML models, data platforms, dashboards and full applications for clients across many industries. We are hiring 5 people; each will lead in 1–2 areas and support the rest of the team.

Responsibilities
• Run AI discovery sessions and turn business problems into AI strategies, roadmaps, PoCs and production MVPs.
• Integrate LLMs (OpenAI, Claude, Gemini, Llama, Mistral) and build NLP features, RAG pipelines, AI agents, agentic workflows and agentic OS setups.
• Set up AI ops, governance and AI gateways: guardrails, evaluation, monitoring, cost and access control.
• Build machine learning solutions: computer vision, predictive analytics and intelligent process automation, deployed with MLOps.
• Design data pipelines (ETL/ELT), data warehouses and real-time analytics; modernize legacy data and enforce data governance.
• Create business intelligence reports and custom dashboards in Power BI, Tableau or Looker.
• Design and develop full-stack products and applications, integrate third-party APIs, and modernize legacy systems.
Required skills and experience
• 6+ years of professional experience in software, AI/ML or data engineering, including 2+ years with LLMs or ML in production.
• Expert Python and/or TypeScript; strong system design, clean code, testing and Git workflows.
• Proven delivery of at least 3 production projects (AI agents, ML models, data platforms or SaaS products) with links or case studies.
• Hands-on with AWS, Azure or GCP, Docker and CI/CD.
• Senior communication: able to lead client calls, estimate work and explain technical trade-offs clearly.

Tools and technologies
• AI & LLMs: OpenAI, Anthropic Claude, LangChain, LangGraph, LlamaIndex, CrewAI, MCP, Pinecone, pgvector, Weaviate, LangSmith
• ML & MLOps: PyTorch, TensorFlow, scikit-learn, OpenCV, YOLO, XGBoost, MLflow, SageMaker, Vertex AI
• Data & BI: SQL, Spark, Snowflake, Databricks, BigQuery, dbt, Airflow, Kafka, Power BI, Tableau
• Product: React, Next.js, Node.js, FastAPI, Django, PostgreSQL, REST, GraphQL, Docker

Expected deliverables
• Working PoCs, MVPs and production features with clean, tested and documented code.
• Architecture diagrams, data models, dashboards and weekly progress updates.

We run an AI detection system on live camera feeds, and we need a Django backend engineer to make the model-serving side faster, more stable, and easier to scale. The detection models already exist. The job is to make the backend that feeds them frames and handles their output run well in production.

What you’ll work on:

Profile and optimize the pipeline from live stream to model inference to results, and remove latency and throughput bottlenecks
Build a frame-handling layer in Django that manages concurrent streams without dropping or backing up frames (threading, multiprocessing, or Celery workers)
Implement frame-skipping, batching, queueing, and backpressure strategies so inference keeps up under load
Build the backend around the detection models: loading, versioning, swapping, and rollback of model versions without downtime
Store and process detection results in PostgreSQL, with careful indexing and query optimization for high-volume event data
Build the Django REST Framework endpoints for detection events, stream and camera configuration, and model status
Add monitoring and logging for latency, FPS, queue depth, and failure rates, plus automatic stream reconnect and recovery
Document the architecture and optimizations you make

Required skills:

Strong Python and Django/DRF, with production experience
Concurrency and performance tuning (threading, multiprocessing, asyncio, Celery)
PostgreSQL query optimization
Redis or a similar queue/cache
Experience profiling and debugging slow or unstable backend systems

Nice to have:

Familiarity with OpenCV and RTSP streams
Experience serving ML models (ONNX, PyTorch, Triton)
Docker and CI/CD

We need to expand firmware development on STM32 for our continuously evolving robots. We are therefore looking for an engineer with experience in STM32, high-speed CAN communication systems, and the management of LiDAR systems and RGB cameras. The duration of the collaboration will depend on the feasibility of conducting on-site testing for any necessary firmware adjustments.

I have an existing race-event platform that already stores user profiles and their photos. The next milestone is to let the system automatically read the bib number on an uploaded image while simultaneously matching the face in that same photo to the correct registered runner. All images will be uploaded manually by users; no live camera feed is involved.

Required behaviour
• When a user submits a picture, the module must detect all athletes wearing a bib and/or face recognition
• If more than one face or bib is found, the photo should be rejected with a concise error message so the user can try again.
• For accepted photos, extract the bib digits (OCR) and save them alongside the file, then confirm the face identity against the existing profile database and surface that person’s details.

Technical notes & deliverables
– Integrate cleanly with the current codebase without breaking existing endpoints.
– Use reliable libraries such as OpenCV, Tesseract, or comparable cloud vision APIs; I am open to your recommendation as long as accuracy is demonstrably high.
– Provide unit tests covering detection accuracy and the single-runner validation rule.
– Include a brief README outlining setup, dependencies, and how the new endpoints or services should be called.

If you have prior work in sports timing, OCR, or face recognition, please share links or repos to speed up selection.

Skills: Python, Django, Software Architecture, MySQL, Face Recognition, OCR, PostgreSQL, Image Processing, OpenCV, Computer Vision

We’re seeking an experienced colorist to deliver polished, consistent color grading for feature films and television series. The role involves refining footage, enhancing visual tone, and maintaining a cohesive look across episodes or films. You should be comfortable working with established creative direction and collaborating closely with directors and editors to achieve the desired final look. Experience with high-profile projects and a strong portfolio of graded content are essential.

Each Episode is 15-20min a piece

Project Domain / Category
Image Processing – Deep Learning / Computer Vision / Web Application
Abstract
Road damage such as cracks is a common problem that can affect traffic safety, vehicle
conditions, and road maintenance. Traditionally, road inspection is performed manually
through physical surveys and visual inspection. These approaches can be time-consuming,
costly, and difficult to perform regularly over large road networks. With advances in computer
vision and deep learning, road images and video can be analyzed automatically to identify
visible road damage.
This project aims to develop an AI-Based Road Damage Detection and Monitoring System using
deep learning techniques. The proposed system will analyze uploaded road images or real-time
video streams to detect and classify road damage such as potholes and cracks. A deep learning
model will be trained using a dataset containing road images with different types of damage.
The model will be integrated with a web-based interface where users can upload images or
provide a live camera feed and view the detected damage with visual indicators. The system
can also maintain detection records and generate reports for monitoring and maintenance
purposes. The proposed technology can support road inspection, transportation management,
and timely maintenance planning.
Functional Requirements
1. User Management & Authentication
Manages user registration, login, and access control for authorized system usage.
2. Image & Video Input
Allows users to upload road images or provide video/camera input for automated road
damage analysis.
3. Road Damage Detection Engine (AI-Based)
Uses a deep learning model to automatically detect visible road damage from images
and video.
4. Crack Detection Module
Detects visible road cracks using image processing and deep learning techniques.
5. Damage Classification Module
Classifies detected road damage into predefined categories such as potholes and cracks.
6. Detection Visualization
Displays detected damage using bounding boxes, labels, confidence scores, or other
visual indicators.
7. Recording & Evidence Management
Stores detected road-damage images, prediction results, timestamps, and related
evidence for future reference.
8. Monitoring Dashboard
Provides a user-friendly dashboard to display detection results, damage categories,
severity information, and system status.
9. Report Generation & Analytics
Generates reports and statistics based on detection history, damage types, severity, and
system performance.
10.Database Integration
Stores user information, detection results, timestamps, image paths, predictions, and
related records using SQLite or MySQL.
11.Admin & Model Management
Allows administrators to manage users, system settings, stored records, and deployed AI
models.
12.Web-Based User Interface
Provides a web-based interface for image uploads, video analysis, viewing detection
results, and accessing monitoring reports.
Tools and Technologies Required
• Programming Language: Python
• Deep Learning: TensorFlow / Keras or PyTorch
• Object Detection / Computer Vision: YOLO, OpenCV, Pillow
• Backend: Flask / Django
• Frontend: HTML, CSS, JavaScript
• Database: SQLite / MySQL
• IDE: PyCharm, VS Code, or Jupyter Notebook
Dataset
• Public Road Damage Detection Dataset (RDD) or a similar publicly available road-damage
dataset
• Road images containing cracks, and other predefined damage categories
• Custom road images (optional) for additional testing or fine-tuning
• Database: SQLite / MySQL

Skills: Python, Machine Learning (ML), Data Science, Image Processing, Tensorflow, Computer Vision, Deep Learning, AI Model Development, AI Development, Data Annotation

We are a growing AI startup seeking a motivated and reliable business partner based in the United States or Canada to assist with the testing and development of our AI-driven market intelligence platform.
Our team has recently developed a computer vision system—built using Python and OpenCV—designed to monitor retail stores, shopping centers, and local markets across the U.S. and Canada to analyze market conditions.
As our platform enters the testing phase, we are looking for a local partner to help with market data preparation, AI model training, system validation, and business development.

Key Responsibilities
- Collect and organize data regarding local markets and the retail sector
- Assist with AI data annotation, labeling, and dataset preparation
- Support the testing and accuracy verification of computer vision models
- Research local retail trends and market conditions
- Provide feedback to improve the performance of AI systems
- Collaborate with the technical team throughout the testing and development process
- Identify opportunities for future business expansion

Qualifications
- Currently residing in the United States or Canada
- Strong communication skills and a systematic approach to tasks
- Basic computer literacy and experience with online research
- Meticulous, highly responsible, and capable of working independently
- An interest in AI technology, market research, or business development
- No prior experience in programming or advanced AI required

What We Offer
- A flexible, remote work environment
- Structured guidance and training from an AI development team
- Hands-on experience in the fields of AI and computer vision
- Potential for long-term collaboration and future business partnership
- Compensation based on the agreed-upon scope of work and responsibilities

We welcome applications from individuals who are passionate about AI innovation, eager to explore local markets, and interested in contributing to a growing technology company.
To confirm that you have read and understood the above requirements, please begin your application with the phrase "BUSINESS PARTNER."

I want to put intelligence behind my existing IP cameras so they can spot unusual activity on their own and trigger the right security responses without me lifting a finger. The core of the job is to build an AI-driven controller that watches the live feed, detects predefined events (motion in restricted hours, unknown faces, package removal, etc.), and then automatically issues the appropriate security alert—push notification, email, or siren—through a simple dashboard or mobile push gateway.

Scope of work
• Connect to standard ONVIF or RTSP cameras on my local network.
• Run real-time video analytics (OpenCV, TensorFlow, or another proven framework is fine) to classify events I can customise later.
• Expose a lightweight web or REST interface so I can arm/disarm zones and review alert logs.
• Push alerts instantly to my phone and optionally trigger a relay (e.g., a smart siren).
• Package everything so it installs on a small box in my office (Raspberry Pi 4 or Ubuntu server—your call, just outline the resource requirements).

Acceptance criteria
1. At least three event types reliably detected (e.g., motion, person, package).
2. <1 second latency from detection to alert on a local network.
3. Clear setup guide that lets me add or remove cameras and tweak sensitivity without touching code.
4. Source code delivered in a Git repo with comments and a brief README.

If you have prior work with Home Assistant, MQTT, or similar automation stacks, please mention it—plug-and-play is a big plus for me.

Skills: PHP, Mobile App Development, Android, Software Architecture, OpenCV, Tensorflow, Computer Vision, AI Automation

We are an independent technology studio seeking an experienced Machine Learning / Computer Vision Engineer to develop a localized Python-based proof-of-concept (POC) script. The objective is to analyze high-resolution 2D photographs of a flexible, non-rigid texturized surface, map a dynamic unit density baseline, and execute dual-method logical matching algorithms based on structural metrics.

The selected developer will be granted secure access to abstract geometric calibration data and texturized surface profiles.

Core Technical Requirements
• Pipeline: Develop a standalone Python script utilizing OpenCV or a lightweight custom YOLO framework executing entirely on-device.
• Frame Normalization: Automatically map, rectify, and flatten perspective tilts and variable ambient lighting from a smartphone camera capture into a uniform 2D coordinate matrix against a physical template.
• Feature Extraction: Project a dynamic, elastic vector grid-mesh over the textured surface to perform automated density verification and precise unit counts per inch.
• Data-Matching Logic: Develop a backend mathematical processor that cross-references the real-world camera count against target data values to output specific configuration and adjustment metrics to the console log.
• Architecture: Code must be fully modular, cleanly documented, and optimized for future packaging inside a mobile SDK/API deployment framework without cloud dependencies.

Milestone & Interview Timeline
We are operating on an accelerated commercial runway and expect strict adherence to the following schedule:
• October 29-November 1 (Wed-Sun): Proposal Review & Shortlisting
• November 2-November 3 (Mon-Tues): 15-Minute Zoom Technical Discovery Calls
• November 4-November 5 (Wed-Thu): Contract Award & Onboarding
• November 13 (Fri): Delivery of Core Algorithmic Logic Parameters

Application Questions (please answer to be considered):
1. What is your experience with projecting dynamic, elastic vector grid-meshes over flexible, texturized surfaces using OpenCV or lightweight YOLO models?
2. How would you handle image rectification to flatten skewed perspective angles and normalize variable lighting conditions into a uniform 2D matrix?
3. Are you available to deliver a clean, modular Python script executing this image-to-console logic pipeline by the November 13 deadline?

We are looking for an experienced AI / Computer Vision Engineer to develop models for anatomical landmark detection using 2D images and 3D scans, primarily for digital dentistry applications.

The engineer will collaborate with our clinical and software development teams to build accurate, reliable, and practical AI solutions.

Requirements
- Strong Python skills and hands-on experience with PyTorch or TensorFlow.
- Proven experience training computer vision models for landmark detection, segmentation, or keypoint detection.
- Experience processing 2D images and/or 3D meshes and point clouds.
- Ability to prepare datasets, establish annotation guidelines, and evaluate model accuracy.
- Familiarity with relevant tools such as OpenCV, Open3D, PointNet, or similar frameworks.
- Experience delivering working AI models and inference pipelines.

### Responsibilities
- Define annotation requirements with our clinical team.
- Prepare and validate datasets.
- Train, test, and improve AI models.
- Evaluate prediction accuracy and identify failure cases.
- Develop AI-assisted annotation workflows.
- Deliver trained models, documented code, and structured landmark coordinates for software integration.

Preferred Qualifications
- Experience in dental AI, medical imaging, or anatomical landmark detection.
- Dental knowledge or previous collaboration with dental professionals.
- Experience with 3D computer vision and CAD software integration.

How to Apply
Please share relevant projects, models you personally trained, GitHub repositories or demonstrations, and your availability.

Dental experience is a plus, not a requirement.
We value strong technical expertise and practical model-development experience.

Work Location: This is an on-site position at our office in Giza, Egypt.

We are looking for an experienced computer-vision engineer to develop a proof-of-concept system for virtual advertising in recorded football broadcast footage.

The initial prototype should replace advertising shown on physical LED perimeter boards with new supplied creative, while keeping the replacement correctly positioned as the broadcast camera moves.

This is an early-stage technical feasibility project. The first version will work with recorded footage rather than a live broadcast feed but we eventually want to get to that.

Initial prototype requirements

Using a short sample of football broadcast footage, the system should:

Identify and track the visible LED perimeter boards.
Replace the existing board content with supplied video or animated graphics.
Maintain correct scale, perspective and positioning during camera movement.
Mask players, officials and other foreground objects when they pass in front of the advertising.
Produce a clean exported 1080p video.
Use automated or semi-automated computer vision rather than manual frame-by-frame editing.
Be capable of being adapted to additional footage and stadium configurations.

The successful freelancer must provide the complete source code, documentation and setup instructions.

Required experience

Applicants should have strong practical experience in several of the following areas:

OpenCV
Python and/or C++
PyTorch, TensorFlow or similar frameworks
Camera calibration
Homography and perspective transformation
Semantic or instance segmentation
Object detection and tracking
Video compositing
FFmpeg or GStreamer
CUDA or GPU optimisation
Real-time or near-real-time video processing

Experience with broadcast graphics, augmented reality, virtual production, sports analytics, Unreal Engine, NDI or SDI workflows would be useful but is not essential.

Initial deliverable

The first paid milestone will involve approximately 30–60 seconds of supplied football footage.

The expected deliverable is:

Replacement of the visible LED advertising.
Stable tracking during camera pan, tilt and zoom.
Foreground masking when players cross the advertising area.
An exported demonstration video.
Complete source code.
Instructions explaining how to run the system with replacement creative.

This is not a video-editing or manual rotoscoping project. We are looking for a technically reusable computer-vision solution.

Potential future work

If the prototype is successful, further work may include:

Longer football broadcasts.
Automated stadium and board calibration.
Virtual 3D advertising carpets.
Support for different stadiums and camera positions.
An operator interface.
GPU optimisation.
Near-live or live video processing.
Multiple regional advertising outputs.

There is potential for a longer-term technical relationship for the right person or team.

Application questions

Please answer the following questions in your proposal:

Describe a relevant project where you anchored graphics or virtual objects to footage from a moving camera.
How would you track the geometry of LED perimeter boards during camera movement?
How would you mask players passing in front of the replacement advertising?
Would you use classical computer vision, machine learning or a combination of both?
How would you minimise manual calibration or rotoscoping?
What processing speed would you expect to achieve at 1080p?
What additional development would be required to make the system operate live?
Which parts of your previous work can you demonstrate on a video call?
Are you personally completing the work, or would any part be subcontracted?

Generic AI proposals will not be considered. Please explain your proposed technical approach and include examples of directly relevant work.

Commercial terms

The project will be structured through fixed-price milestones.

The client must receive:

Ownership of all bespoke source code following payment.
Access to the code throughout development through a client-controlled repository.
A list of all third-party libraries, models and licences.
Full disclosure of any existing proprietary components.
Documentation sufficient for another engineer to run and review the system.

The supplied footage and commercial information must remain confidential and may not be used in portfolios or demonstrations without written permission.

Please include:

Your proposed budget for the initial prototype and full development.
Estimated delivery schedule.
Recommended technology stack.
Any assumptions or technical limitations.

Project Domain / Category
Artificial Intelligence / Computer Vision / Medical Image Processing / Deep Learning
Abstract / Introduction
Brain tumor detection from Magnetic Resonance Imaging (MRI) is an important medical image
analysis task. Deep learning techniques can assist in identifying abnormal patterns in brain MRI
images. Capsule Networks (CapsNet) offer an alternative to conventional Convolutional Neural
Networks (CNNs) by learning spatial relationships among image features. Integrating a trained
deep learning model into a web-based application can make MRI image analysis more
accessible and convenient for demonstration and educational purposes.
This project aims to develop a web-based brain tumor detection system using a Capsule
Network (CapsNet). The publicly available Br35H Brain MRI dataset, containing Tumor and NoTumor images, will be used for model training and testing. The developed web application will
allow users to upload MRI images, perform basic image-quality assessment, and obtain Tumor
or No Tumor predictions using the trained CapsNet model.
Functional Requirements: In this project, we will leverage image processing and deep learning
techniques to address the following objectives:
1. Data Collection and Preprocessing: Gather a diverse dataset of brain MRI scans,
ensuring data quality and integrity. Preprocess the images to enhance their suitability
for further analysis. Apply image normalization, resizing, and suitable data
augmentation.
2. Dataset Splitting: To facilitate model training and evaluation, we will split the dataset
into distinct sets for training, validation, and testing. This step ensures that the deep
learning model's performance is rigorously assessed and prevents overfitting.
3. Deep Learning Model Development: Implement and train a conventional Convolutional
Neural Network (CNN) as a baseline model and a Capsule Network (CapsNet) as the
primary model for classifying brain MRI images into Tumor and No Tumor categories.
Conduct a comprehensive comparative performance analysis of both models using
Accuracy, Precision, Recall, F1-Score, ROC-AUC, and Confusion Matrix. The comparison
shall evaluate their classification effectiveness, ability to correctly identify tumor cases,
and computational efficiency to determine the relative strengths and limitations of
CapsNet compared with the conventional CNN.
4. Model Evaluation: We will establish robust evaluation metrics to assess the
performance of the deep learning models, ensuring that the segmentation results are
precise and reliable.
Evaluate the model using Accuracy, Precision, Recall, F1-Score, ROC-AUC, and
Confusion Matrix.
Compare CapsNet with a basic CNN model to determine whether Capsule Networks
provide any performance advantage. Analyze the performance drop of each model
under different image-quality degradations.
5. DataSet: For this project you need to use BR35H Brain MRI dataset.
6. Model Deployment: Students shall develop a user-friendly web interface that allows
users to upload brain MRI images. The uploaded MRI image shall be processed using the trained CapsNet model. The system shall classify the image as Tumor or No Tumor and
display the predicted class and model score

Skills: Python, Image Processing, OpenCV, Flask, Tensorflow, Keras, Pytorch, Computer Vision, Deep Learning, Web Application

Job Title: AI Automation Specialist – Vehicle Rental Business

I run K.A Rentals Ltd, a UK vehicle rental business, and need a freelancer to build an AI assistant that monitors my fleet and automatically reminds me of upcoming tasks and compliance deadlines.

We currently use Fleetworks, DocuSign, Excel spreadsheets and reminders. I want these connected wherever possible to reduce manual checks and missed deadlines.

Main requirements:

* Monitor whether vehicles show as insured (“green”) on the Motor Insurance Database (MID), using an authorised method where available, and flag missing or unverified status.
* Track MOT, road tax, insurance renewals, servicing and other relevant vehicle compliance dates.
* Send advance reminders and urgent alerts through WhatsApp, email or SMS.
* Flag missing documents, unsigned DocuSign agreements and overdue tasks.
* Provide a simple dashboard showing each vehicle’s status, upcoming deadlines and outstanding actions.
* Escalate unresolved issues and keep a record of checks and reminders.

Ideal freelancer:

Experience with AI assistants, workflow automation, APIs and tools such as Make, Zapier or n8n. Experience with UK fleet or vehicle rental systems would be helpful.

Deliverables:

A working, tested system, straightforward setup instructions and a handover session. The system must protect customer data and clearly flag failed checks.

Please include relevant examples, your proposed approach, any integration limitations, estimated delivery time, setup cost and ongoing running costs.

Project Overview

I’m developing a beauty technology startup focused on making professional-looking makeup accessible and affordable at home.

The concept is an app that allows users to upload a photo of themselves, preview different makeup looks virtually, and eventually receive a personalised makeup kit with the products and a tutorial needed to recreate their chosen look.

I’m looking for an experienced developer to build an initial working prototype of a custom virtual makeup try-on tool featuring lipstick, blush and eyeshadow.

This is the first stage of a larger project. I want to establish a reliable, realistic makeup-rendering foundation that can be expanded into a full AI-powered beauty application.

I’m open to adapting suitable existing technology or open-source projects, provided the solution is suitable for commercial use and can be customised to meet my requirements.

Scope of Work

1. Photo Upload

* Allow users to upload a portrait photo in JPG, PNG or WEBP format.
* Automatically detect the face and relevant facial features.
* Display the original photo and the edited result.
* Include basic error handling for unsuitable images.

2. Virtual Lipstick Try-On

* Apply different lipstick shades, including nude pink, rose, classic red, berry and brown nude.
* Include an adjustable intensity slider.
* Preserve natural lip texture, highlights, shadows and lip shape.
* Accurately identify the outer and inner lip boundaries.
* Avoid colouring teeth or the inside of the mouth.

3. Virtual Blush Try-On

* Apply blush naturally to the cheeks.
* Offer several colours, such as soft pink, peach and rose.
* Allow users to adjust the intensity and placement.
* Blend the colour naturally into the original skin while preserving skin texture and facial features.

4. Virtual Eyeshadow Try-On

* Offer a selection of eyeshadow colours and preset styles, such as natural brown, soft pink and evening glam.
* Accurately detect the eyelids and apply colour within the appropriate areas.
* Allow users to adjust the intensity.
* Preserve the natural appearance of the eyes and surrounding skin.

5. Complete Makeup Looks

Users should be able to combine lipstick, blush and eyeshadow to create a complete makeup look.

The prototype should include a few preset combinations, allow users to switch between them, and provide a before-and-after comparison.

6. User Interface

* Clean, modern and beauty-brand-appropriate design.
* Responsive layout for desktop and mobile browsers.
* Simple controls for selecting colours, styles and intensity.
* Ability to reset the image and download the finished result.

Technical Requirements

I’m open to your recommendations on the best technical approach. Potential technologies include:

* React and TypeScript.
* MediaPipe or another suitable facial landmark detection solution.
* Canvas, WebGL or other appropriate image-rendering techniques.
* Image segmentation or additional computer vision techniques where necessary.

You may propose adapting an existing open-source project or SDK if it offers a suitable foundation.

However, please disclose any third-party dependencies, licensing restrictions, commercial-use permissions, ongoing costs and technical limitations before implementation.

The main priority is realistic makeup rendering, not simply applying flat blocks of colour to the face.

The finished result should blend naturally with the user’s features and preserve the original person’s identity and appearance as much as possible.

Privacy and Data Handling

Where practical, uploaded photos should be processed locally in the browser. If external processing or third-party APIs are required, please explain what data is transmitted, where it is processed, and any associated costs.

Deliverables

1. A working web-based prototype featuring lipstick, blush and eyeshadow.
2. Several preset combinations demonstrating complete makeup looks.
3. Complete source code and project files.
4. Instructions for running, testing and maintaining the application.
5. A demonstration using a variety of suitable portrait photos.
6. Documentation of third-party libraries, licences, ongoing fees and limitations.
7. Confirmation of ownership and usage rights for custom code and project-specific deliverables, subject to any disclosed third-party licence terms.

Future Development

If the prototype performs well, I intend to explore additional features, including:

* AI-generated personalised makeup looks.
* Recommendations based on occasion, time of day, outfit, preferred style and skin concerns.
* More makeup categories and realistic product-specific shades.
* Live camera try-on.
* Integration with a product recommendation and makeup-kit ordering system.
* A future mobile application for iOS and Android.

These features are not required for the initial prototype, but the solution should be reasonably extensible.

Ideal Candidate

I’m looking for someone with demonstrable experience in one or more of the following:

* Computer vision and facial landmark detection.
* Image processing and realistic cosmetic rendering.
* Canvas or WebGL-based image manipulation.
* Virtual makeup try-on or beauty technology.
* Building functional prototypes with clean, maintainable code.

Please include examples of relevant projects you have personally worked on.

Budget and Proposal

My initial budget is for the prototype, depending on the proposed approach, existing technology that can be reused, and the quality of the results.

I understand that a highly polished commercial product may require a larger budget. I am therefore open to a staged approach, beginning with a clearly defined milestone and expanding the scope once the initial results have been evaluated.

In your proposal, please explain:

1. How you would approach building this prototype.
2. Whether you would use existing technology or develop the rendering functionality from scratch.
3. How you would achieve realistic blending for lips, cheeks and eyelids.
4. Examples of similar projects you have completed.
5. Your estimated timeline and fixed-price quote.
6. What is included in your price and what would incur additional costs.
7. Any limitations or compromises you anticipate within the proposed budget.

I value technical quality, realistic results and a maintainable foundation over simply delivering a large number of features.

I would prefer to start with a small milestone to validate the rendering quality before proceeding with the full prototype.

I need a custom virtual camera that can join a Zoom call and replace my face in real time with a chosen subject’s face. Before each meeting I want to load one or more reference photos (or short video clips) of that person; once the call starts, the software should track my head and mouth movements and render a realistic, latency-free face swap that other participants see as my camera feed.

Zoom is the first platform I must support, but I eventually want to point the same virtual camera at Google Meet and Microsoft Teams, so please keep cross-platform thinking in mind. Whatever stack you prefer—OpenCV, MediaPipe, TensorFlow, Unreal’s LiveLink, or your own pipeline—you’ll need to handle live video I/O, facial-landmark tracking, texture blending, and create a signed driver or plugin so the operating system recognises the feed as a standard webcam.

Deliverables
• Desktop application (Windows or macOS) that appears in Zoom as a selectable camera
• Interface to import photos or videos that train/drive the swap model on the fly
• Real-time, high-resolution face replacement that matches lighting and expressions
• Installation package and brief usage guide for non-technical users

When you reply, just show examples of past work that prove you have already built—or heavily contributed to—similar real-time video or AR projects. A short demo clip or repo link is perfect.

Skills: C Programming, C# Programming, C++ Programming, OpenGL, Augmented Reality, Artificial Intelligence, OpenCV, Tensorflow, Computer Vision, Facial Recognition

I’m developing an AI-powered loss-prevention system for delis and convenience stores.

I own a deli in Brooklyn, NY, and want to build an MVP using our existing security cameras. The goal is to use computer vision to identify potentially suspicious activities involving cashiers and customers, such as unusual register activity, potential employee theft, and shoplifting.

The system should ideally:

* Connect to existing IP cameras and DVR/NVR systems using RTSP or ONVIF where supported.
* Analyze video footage using AI.
* Detect and flag specific suspicious events.
* Save short video clips with timestamps for review.
* Provide a simple dashboard and alerts for business owners.
* Eventually support multiple cameras and multiple business locations.

I’m looking for an experienced computer vision engineer with a background in CCTV, real-time video analytics, and AI surveillance systems.

I want to start with a small technical feasibility project using my own store’s cameras before investing in a full commercial platform.

I’m interested in your recommended technical approach, estimated timeline, and budget for an initial MVP.

The goal is to build a reliable, scalable product that can eventually be offered to other small businesses through a monthly subscription.

I’m looking for an experienced AI computer vision developer to help build an MVP for an AI-powered loss-prevention service for independent grocery stores and super markets.

The concept:

* Analyze footage from existing CCTV cameras and DVR/NVR systems.
* Detect potentially suspicious cashier behavior, such as concealing cash, suspicious handoffs, or unusual product handling.
* Send timestamped alerts and video clips to the store owner.
* Work without requiring a POS system.
* Avoid replacing existing cameras wherever technically possible.

Required experience:

* Computer vision and video analytics.
* Python, OpenCV, and relevant AI video models.
* RTSP, IP cameras, and DVR/NVR integrations.
* Real-time video processing and alert systems.
* Deploying AI locally on an edge device or in the cloud.

First milestone:
I want a technical feasibility assessment and a small proof of concept using sample deli footage before committing to a full product.

Please share examples of similar systems you’ve built, your proposed technical approach, estimated cost, and timeline.

We are looking for an experienced Computer Vision & AI Software Engineer to develop, fine-tune, and deploy a custom real-time object detection system using YOLOv8.
The goal of this project is to build an end-to-end solution—from fine-tuning a YOLOv8 model on custom datasets to deploying a lightweight backend API and Web interface for live inference (image/video/webcam streams).
Key Responsibilities
• Data Preparation & Fine-Tuning: Preprocess, augment, and split custom image/video datasets to fine-tune a YOLOv8 model (e.g., target classes such as PPE compliance: Hardhat, No-Hardhat, Safety Vest, No-Safety Vest, Mask, Person, etc.).
• Model Training & Evaluation: Train and evaluate the model using PyTorch and Ultralytics YOLOv8. Monitor loss curves and report metrics including [email protected], [email protected]:0.95, Precision, and Recall.
• Backend API Development: Build a clean FastAPI backend endpoint (/predict) to handle image upload / video frame stream requests and return inference results with bounding boxes (e.g., encoded as Base64 or JSON bounding box formats).
• Frontend Web & Live Stream Integration: Create a simple HTML/JavaScript frontend supporting image uploads, local video file processing, and live webcam stream detection.
• Model Optimization & Export: Export the trained model to production-ready formats (e.g., ONNX, TensorRT, or .pt) and document deployment steps.
Required Skills & Qualifications
• Machine Learning / CV: 2+ years experience in Object Detection, Deep Learning, and Computer Vision (OpenCV, PyTorch, Ultralytics YOLOv8).
• Backend Development: Strong proficiency in Python and FastAPI.
• Frontend Basics: Basic understanding of JavaScript/HTML for handling web assets and canvas bounding box rendering.
• Model Evaluation: Ability to deliver detailed evaluation reports (Confusion Matrix, PR curves, F1 curves).
Deliverables
1. Trained Weights & Model Checkpoints: Optimized .pt and ONNX weights hosted or uploaded to Hugging Face / GitHub releases.
2. Training Artifacts & Notebook: Annotated Jupyter notebook (YOLO_Training.ipynb) showing dataset prep, loss curves, confusion matrix, and evaluation metrics.
3. End-to-End Application Codebase:
o Backend (backend/app.py): FastAPI inference server.
o Frontend (index.html, webcam.html): Web interface for upload and real-time webcam monitoring.
o Local Script (local/webcam_detection.py): Python script for local live webcam stream testing.

……………………………………………………………………………………………………………………………………

I’m an independent artist creating dark, melodic R&B inspired by The Weeknd and Giveon. I’m looking for a remote session operator to support my recording process while I focus on singing, freestyling, and songwriting. You’ll load beats into FL Studio, start and stop vocal recordings, play back takes, handle punch-ins and retakes, organize vocal files, and save projects. Sessions will be conducted online through Discord or a similar platform, with FL Studio running on my computer. I want someone patient, reliable, nonjudgmental, and easy to work with. I’d like to start with a paid 30–60 minute trial session.

We are seeking an experienced Computer Vision Engineer to build an end-to-end crack detection and mensuration pipeline. The objective is to produce a model that can identify, segment, and quantitatively measure crack dimensions across the following structural surfaces: concrete, brickwork, mortar joints, and varied masonry.
Expected job functions are:
1. Scrape publicly available datasets containing cracks representative of what is seen on masonry structures
2. Combine publicly available datasets with proprietary datasets from aRoboticsCompany
3. Label and segment relevant cracks in those datasets (relevance to be informed by aRoboticsCompany team)
4. Select an appropriate pre-trained model to conduct transfer training on
5. Conduct transfer training, applying labeled crack dataset to that model
Model expected functions:
1. Generate a semantically segmented mask over cracks in an image
2. Generate estimated length and width for cracks in mask
3. Generate JSON’s or XML’s that list each crack location and metrics in each image

Dataset curation (expected 30 hours of work)
Identify, scrape, and clean publicly available datasets containing masonry/brick/mortar cracks (e.g., open academic datasets, structural inspection archives, web sources) while ensuring quality control and licensing compliance.
Merge curated public data with our internal proprietary crack datasets. Establish standardized annotation formats (COCO-stuff , binary masks, or polygon GeoJSON).
Implement tiling/slicing workfl ows (e.g., SAHI / patch-based inference) to preserve micro-crack details in high-resolution images, along with robust data augmentation (lighting shifts, mortar texture noise, weathered brick surfaces).

Data labeling requirements (expected 200 hours of work)
Label cracks in dataset in accordance with aRoboticsCompany requirements. Labels must consist of manually generated shading or masks over cracks in the training images with pixel level accuracy.

Model training (expect 20 hours of work)
Train, validate, and benchmark high-resolution segmentation models specialized in thin, elongated topologies (e.g., SegFormer, HRNet, UNet with attention/transformers, or YOLOv8/v11-seg).
Ensure the model generalizes across surface variations: differentiating actual structural cracks from natural mortar lines, surface spalling, shadows, effl orescence, and vegetative growth.
Implement post-processing algorithms (e.g., medial-axis skeletonization, Distance Transform, Voronoi/graph-based centerline extraction) to calculate:
Crack Length: Continuous path tracking along the crack skeleton.
Crack Width (Aperture): Local orthogonal width profi les along the skeleton (mean, median, and maximum width).
Model backtesting and documentation (expect 24 hours of work)
Provide clean, well-documented training and data-scraping codebase (PyTorch preferred)
Provide backtesting data quantifying model accuracy

Final Deliverables:
Standalone inference pipeline accepting an input image and returning:
High-resolution semantic crack mask.

Overlay visualization highlighting detected cracks.

Structured output (JSON) reporting crack instance ID, coordinates, estimated length, and estimated width profi le for each named image.

Brief technical report summarizing validation metrics (mIoU, F1/Dice score, width/length error margins against reference test sets).

O que contratar primeiro

**Especialista em visão computacional para futebol, rastreamento de múltiplos jogadores e reconhecimento temporal de ações.** Procurar experiência demonstrável em Python, PyTorch, OpenCV, detecção de objetos, associação de trajetórias e avaliação de vídeo. Conhecimento de YOLO/BoT-SORT e reidentificação por aparência pode facilitar o aproveitamento da base; trocar componentes exige comparação e justificativa.

Não exigir uma LLM própria ou um modelo generativo como solução inicial. O problema atual envolve detecção, associação temporal, identidade, bola e classificação. A arquitetura pode combinar modelos especializados. O produto deve executar a análise sem chamadas ao Gemini/OpenAI ou a outro serviço de compreensão visual; hospedagem em GPU é infraestrutura substituível.

## Base que deve ser reaproveitada

| Ativo existente | O que oferece | O que ainda falta |
|---|---|---|
| Detector de pessoas, rastreador e pose | Observações de vários jogadores, caixas e IDs locais | Continuidade confiável em sobreposição, saída/retorno e cortes |
| Detector próprio de bola refinado por 42 épocas | Caixas e observações da bola em vídeos de celular | Cobertura em bola pequena, borramento, oclusão e bola aérea; bola detectada não comprova posse |
| Classificador de ações e refinamentos anteriores | Propostas de passe, condução, finalização e outras classes | Reduzir falsos positivos e ações perdidas; ligar ação ao jogador correto |
| JSONs auxiliares e revisões humanas | Supervisão já aproveitada em treinos próprios | Não são trajetórias densas revisadas de todos os jogadores nem de remetente/receptor |
| Corpus histórico de ações | Dez fontes, 49 positivos e 15 negativos específicos; versão agrupada com 70 grupos | Poucos exemplos e fontes por classe; 70 grupos não são 70 vídeos independentes |
| Contexto e índice de participantes | Evidências por frame e propostas ligadas a atores possíveis | Tracklet não equivale a pessoa global ou perfil; autoria e recepção ainda ambíguas |
| Worker Python e contrato TypeScript | Processamento de pesquisa e JSON estruturado | Integração durável ao upload e aceite de produto são entregas posteriores |
| Backup local | Pesos, corpus, revisões, relatórios e configurações preservados | Fornecer pacote mínimo selecionado; não repassar todo o backup |

Os 11.327 rótulos de bola registrados no intake pertencem à tarefa de bola, não são 11.327 ações. As revisões de identidade disponíveis são esparsas e não substituem anotação de trajetórias. Quantidades finais elegíveis e linhagem devem ser conferidas no inventário do primeiro marco.

## O que precisa de treino e o que precisa de engenharia

| Frente | Trabalho necessário | Prioridade |
|---|---|---|
| Identidade e rastreamento | Diagnosticar associação, estado durante oclusão e recuperação; avaliar se embeddings de aparência precisam de adaptação com pares/identidades anotados | Primeira contratação |
| Bola | Avaliar o checkpoint existente por tipo de erro; se necessário, refinar com caixas revisadas de bola pequena/aérea/oculta e negativos explicitamente anotados | Condicionada ao erro medido |
| Contato e transição entre jogadores | Anotar/evaluar jogador que toca, instante, bola livre, receptor e incerteza; produzir candidatos sem depender apenas do atleta inicial | Segunda etapa |
| Ações temporais | Refinar passe, condução e finalização com clipes/intervalos revisados, máscaras de desconhecido e negativos por classe; avaliar gol separadamente | Segunda etapa |
| Autoria e contagens | Associar evento, autor e receptor a IDs reconciliados; separar tentativa, conclusão e resultado desconhecido | Após evidência de identidade e ação |
| Backend e melhoria contínua | Versionar dataset/modelo, registrar inferência, preservar revisões e avaliar candidatas antes de promoção | Entrega de reprodução; integração do app orçada à parte |

Falta, desarme, drible contra adversário e gol não devem herdar automaticamente o resultado de uma classe parecida. Condução não prova drible bem-sucedido. Bola perto do jogador não prova domínio. Finalização não prova gol. Um gol fora de quadro pode ficar desconhecido.

Skills: Python, Data Processing, OpenCV, Pytorch, Computer Vision, Deep Learning, Object Detection, YOLO, Data Annotation, Model Evaluation

I need a Windows desktop application built with Python 3.12 that connects to a Hikrobot industrial camera through the MVS SDK, runs a custom YOLO (Ultralytics) model via OpenCV/NumPy, and reliably detects individual metal bars scattered on a work surface. Any orientation—horizontal, vertical, diagonal, or arbitrarily rotated—must be handled, and when bars overlap the software should still outline each one separately and decide which bar sits physically on top by analysing the single 2-D image alone (no depth sensors or special labels).

While the camera is streaming the GUI should:
• show the live image with coloured outlines around every detected bar, each overlay carrying a unique ID number;
• display, in a side panel or table, the ID, 2-D position, rotation angle, and a “Top / Below” indicator that states whether that bar is above or underneath another.

Deliverables
1. Executable Windows desktop application and all source code.
2. Trained YOLO weights (with brief notes on the training dataset so I can retrain later).
3. Setup and run instructions that cover Python 3.12 environment creation, required pip packages, and Hikrobot MVS SDK integration.
4. A short test video or image set demonstrating correct detection, orientation output, and top-bar identification.

Acceptance criteria
• Detection accuracy remains stable when bars are rotated to any angle or partially occlude one another.
• GUI refreshes in near real-time on a typical industrial PC.
• Top/Below flag matches visual inspection on at least 95 % of overlapping cases in the provided test set.

Skills: Python, Windows Desktop, Software Architecture, Machine Learning (ML), C++ Programming, Image Processing, OpenCV, NumPy, Computer Vision, YOLO

About the project:
We have a working edge AI analytics application on Raspberry Pi 5 with a Hailo-8 accelerator. It runs computer vision on-device and outputs only anonymous counts and metrics, never video. The application is built and tested:

C++17/20 inference binaries (HailoRT, GStreamer, OpenCV)
A C++ process supervisor that manages the inference processes
A Python service: FastAPI, MQTT, SQLite, scheduling
Unit and integration tests, plus mock builds that run without hardware

What we don't have is the production platform around it. Right now devices are set up by hand. We need someone experienced in embedded Linux to turn this into something we can install, update and support in the field without touching each device.

Scope of work:
Reproducible device image. Build a Raspberry Pi OS (64-bit) image with one command. It should include a pinned kernel, the HailoRT driver and runtime, our packaged applications, and models. Tooling could be rpi-image-gen, pi-gen, or Yocto if you can justify it.
Application packaging. Package our C++ and Python components as .deb packages or another standard format, using standard filesystem paths and systemd units. We'll do much of this ourselves first; we need you to review it and standardize it.
Safe over-the-air updates. A/B system updates with automatic rollback triggered by our health endpoint, signed artifacts, and separate app-only updates. We expect Mender, but we're open to RAUC or SWUpdate with a reasoned recommendation.
Handling the HailoRT kernel driver. A strategy for updating the kernel and OS without breaking the HailoRT kernel module.
Device hardening. No default credentials, non-root services, a firewall, an SSH policy, a read-only or overlay root filesystem where practical, a hardware watchdog, and protection against power loss and SD card wear.
Provisioning and identity. Unique per-device credentials or certificates, plus a first-boot setup flow for device name, site and network.
CI pipeline. Automated ARM64 builds, versioned releases and artifact signing. GitHub Actions or GitLab CI is preferred.
Documentation. A runbook covering how to build, release, provision a new device, roll out an update, and recover a failed device.

Milestones (fixed price per milestone):
M1: A reproducible image that boots and runs our application, built from a script in our repo.
M2: Working OTA updates, including a live demo where a deliberately broken update rolls back automatically.
M3: Hardening, provisioning and per-device identity complete.
M4: CI pipeline and runbook delivered and handed off.

Required experience:
Shipping embedded Linux products to the field, not only servers or cloud
Raspberry Pi or other ARM single-board computer image builds (pi-gen, rpi-image-gen, Yocto or Buildroot)
OTA systems with A/B partitions and rollback (Mender, RAUC, SWUpdate)
systemd, Debian packaging, kernel modules / DKMS
Linux security hardening for unattended devices

Nice to have:
Hailo, NVIDIA Jetson, or other edge AI accelerators
Camera and GStreamer pipelines on embedded Linux
Fleet management platforms (Mender, balena)

Not in scope:
Application logic, inference models, the API, or dashboards. Those stay with our team.

To apply, please answer:
How would you make updates fail-safe on a device mounted on a ceiling with no physical access?
How do you handle kernel updates when a vendor kernel driver (like HailoRT) depends on a specific kernel version?
How would you give each device unique credentials during provisioning?
Describe an embedded device fleet you shipped: hardware, update system, and number of devices.

An NDA and IP assignment agreement will be required before you get access to the repository.

Need an expert to implement robust object tracking on the imx8m platform. The solution should handle occlusion recovery, support zoom in/out, and adapt to shape changes during tracking. Prior experience with similar tracking implementations is required. Please share relevant examples of past work and your approach to building a reliable tracking system for this platform.

Hiring: AI Automation Developer – Automated Apparel Mockup System

Job Type: Remote / Freelance / Project-Based
Compensation: Based on experience and project proposal

Project Description

seeking an experienced devoper to build an automated clothing mockup generation system using AI-assisted image processing, Python, Photoshop automation, or other suitable technologies.

We manufacture our own clothing and already have professional product photographs featuring real models wearing our blank garments.

We do NOT need AI-generated models or garments.

Our goal is to automatically apply new print graphics to our existing model photographs while preserving the original appearance of the model and garment.

Required Workflow

1. Upload Print Designs

Our team uploads new print graphics into a designated Google Drive folder or server.

2. Automatic Processing

The system detects new designs and automatically retrieves our existing clothing mockup templates.

3. Apply Graphics to Existing Model Photos

The system must:

- Keep the original model's face, body, hair, skin, pose, and appearance completely unchanged.
- Keep the original garment's shape, fit, wrinkles, seams, fabric texture, and proportions intact.
- Preserve the original background, lighting, and shadows.
- Apply the new graphic realistically onto the garment.
- Automatically adjust print placement, scaling, perspective, and distortion to follow the garment's natural shape.
- Preserve the exact original artwork, colors, typography, and design details.
- Support multiple garment colors, styles, and print positions.
- Prevent graphics from appearing on skin, hair, or background.
- Generate consistent results across different designs using the same clothing template.

Important: The system must edit only the designated print area of the existing photograph. It must not regenerate or alter the model or garment.

4. Batch Mockup Generation

For example:

One uploaded design should automatically generate mockups using multiple existing Phillove clothing templates, such as T-shirts, sweatshirts, tank tops, and other garments.

5. Automatic File Saving

Completed mockups should automatically be saved in organized folders with filenames identifying the design, garment style, color, and view.

Preferred Technology

We are open to:

- Python, OpenCV, and image compositing
- Adobe Photoshop Smart Object automation
- n8n or Make.com workflow automation
- AI-assisted garment masking and image processing
- Google Drive API integration

Accuracy, consistency, automation, and preservation of our original photographs are more important than using generative AI.

Expected Deliverables

1. Fully automated mockup generation system.
2. Automatic Google Drive folder monitoring.
3. Integration with our existing real-model clothing photographs.
4. Accurate graphic placement and realistic fabric integration.
5. Support for multiple garment styles and colors.
6. Batch processing and automatic file organization.
7. High-resolution e-commerce-ready output.
8. Error handling and quality control.
9. Complete source code, documentation, and training.

Required Test Before Final Approval

We will provide sample model photographs and print designs.

The developer must demonstrate that the system can automatically produce realistic mockups without changing the model, garment construction, or original artwork.

The system should also be tested with multiple designs on the same garment template to demonstrate consistency.

How to Apply

Please include:

- Examples of automated apparel mockup systems you have developed.
- Your proposed technical approach.
- Whether you can preserve the original model photograph without AI regeneration.
- Estimated project cost.
- Estimated completion time.
- Expected monthly operating expenses.
- Details of any ongoing software subscriptions.

We are hiring a developer to build an automated system, not a graphic designer to manually create individual mockups.

About Cranium Inc. (NOEMA)
Cranium Inc. is building a wearable platform that combines computer vision, AI, and AR/VR. We are building towards a functional prototype and are looking for an experienced engineer to take over technical leadership and carry it forward.

The role
Our current lead engineer is stepping down for health reasons, effective October 31, 2026. You'll take over his work through a structured handoff, supported by NOEMA, our engineering assistant that holds the project context, codebase knowledge, and documentation. From there, you'll own the continued development of the prototype.

What you'll do

-Pick up the existing prototype work and get up to speed quickly through the handoff
-Build and train computer vision detection models that find and classify what the wearable sees in real time, such as people, vehicles, and objects, so the system knows both what is in view and where it is
-Lead development of the computer vision and AI pipelines running on the wearable
-Build and refine AR/VR features and the user-facing experience
-Design and integrate agentic AI capabilities into the platform
-Write production-quality C++ and Python, and make the key architecture decisions
-Document your work and set the technical direction for the next phase

Requirements

-PhD or Master's in Computer Science, Computer Engineering, or a closely related field
-Strong expertise in computer vision (detection, tracking, SLAM, sensor fusion, or similar)
-Proven experience with AI/ML, including agentic AI systems (LLM-based agents, tool use, autonomous workflows)
-Hands-on AR/VR development experience
-Advanced proficiency in C++ and Python
-Able to inherit an existing codebase and deliver independently

Nice to have

-Experience with wearables, embedded or edge devices, or real-time systems
-Familiarity with Unity, Unreal, OpenXR, OpenCV, PyTorch, or TensorRT
-Experience in hardware/software prototyping or at a startup

Timeline
We'd like to start right away so the handoff overlaps with the current lead before October 31, 2026.

I’m the founder of WorkCam Secure and need help diagnosing why two physical cameras—WorkCam M4 and Cammpro S4A (S911609)—aren’t connecting and sending video into my existing Lovable + Supabase platform. The setup includes a DigitalOcean server running MediaMTX and Mux integration. I need someone to troubleshoot the current setup, identify the exact failure, and get at least one camera streaming into the dashboard. I want a working demonstration with the actual camera, not a simulated stream, and I want to preserve the existing platform and source code. Please share your experience with physical RTMP/RTSP camera integration, MediaMTX, Mux, and Supabase, and confirm you can diagnose the issue using server logs and network tests.

I work with MMPI answer sheets that have been photographed with a phone. Each sheet shows one question per line with “Evet / Hayır” options, and the correct option is simply marked with an “x”.

What I need is a lightweight program that can:
• load a batch of these photos (JPG or PNG),
• detect every question line, recognise whether the “x” sits under Evet or Hayır,
• export a clean .xlsx file that lists the question number in one column and the detected correct answer in the next.

Accuracy is critical—there should be no manual correction after export. If you choose to use Python, OpenCV and Tesseract that is fine, but any stack is acceptable as long as it runs on Windows 10 without extra paid licences.

Deliverables
1. Executable or script with brief setup notes.
2. Sample Excel output generated from three supplied photos so I can validate detection quality.
3. Well-commented source code.

Acceptance criteria: on the test set I supply, every line must be read and transferred to Excel with 99 %+ accuracy; rows should stay in their original order.

If any part of the spec is unclear, let me know before you start so we can lock it down together.

Skills: Python, Data Processing, Excel, Software Architecture, Software Development, Programming, Image Processing, Data Extraction, Data Analysis