Find OpenCV Projects on Upwork with Vollna

I am developing an indoor research-and-development rig that will rely on up to fifteen miniature Wi-Fi, IR-capable cameras, each fitted with a wide-angle lens. Every feed must reach a single on-site server with minimal latency and rock-solid stability.

Here is what I need from you:

• Help me shortlist and procure the right camera modules, lenses, IR illuminators, access points, and server-side components.
• Design the network topology so all streams coexist on 2.4 GHz or 5 GHz without congestion.
• Recommend or build the software stack—whether that’s RTSP, WebRTC, or a custom pipeline—that can ingest, display, and record all fifteen feeds simultaneously.
• Walk me through the full configuration: firmware flashing, stream settings, server setup, and storage management.
• Remain available during initial trials to troubleshoot connectivity glitches, frame drops, or thermal issues until the whole array runs uninterrupted for 48 hours.

Tools and protocols you’re comfortable with—FFmpeg, GStreamer, OpenCV, ONVIF, Linux, or similar—will be a major plus. Final outcome is a fully documented, working system ready for R&D data collection.

We are building an MVP for a SaaS platform that processes technical documents and generates structured outputs to support workflow automation.

We are looking for an experienced developer with strong background in:

Computer Vision (image/PDF processing)
Python (preferred)
Experience working with structured or semi-structured documents
Cloud platforms (Azure preferred but not mandatory)
Scope of work:
Process PDF documents
Detect and classify predefined visual elements
Extract structured data
Generate CSV outputs
Implement a simple workflow (upload → process → output)

The system should also support confidence-based outputs, allowing identification of elements that may require user validation.

This is a focused MVP with potential for longer-term collaboration.

Please include examples of similar work and a brief outline of your approach.

• How would you approach detecting and extracting structured data from complex PDF documents containing repeated visual elements?
• Have you worked on any projects involving computer vision on technical or structured documents? Please provide examples.
• What tools, libraries, or frameworks would you use for this project and why?
• How would you handle uncertainty or low-confidence detections in the output?
• What would be your estimated timeline to deliver a first working MVP for this type of system?

I am building a Windows-only desktop automation tool in Python that watches the screen, recognises visual cues, and then acts with believable, human-like timing. Using OpenCV (or a comparable vision library) alongside PyAutoGUI, the script should reliably detect Buttons, Icons, and some custom Map Elements that I will supply during onboarding. Once a target is recognised, the script must respond with Mouse clicks, Keyboard inputs, Scrolling, and Drag and Drop actions, each executed through slightly randomised delays and curved cursor paths so nothing looks synthetic.

The application will run quietly in the background, so please structure the code to avoid stealing focus or locking the UI. Prior experience automating full-screen desktop apps will be valuable; if you have tackled game or GIS interfaces before, even better.

Deliverables
• A well-commented Python 3 script (or small module set) that runs on Windows without manual intervention
• Image templates or a training routine for the Map Elements, packaged in an orderly folder structure
• A short README explaining configuration options, how to add new targets, and how to tweak delay profiles
• Optional: a minimal CLI or tray icon to start/stop the automation

Acceptance criteria
• >95 % detection accuracy on my supplied sample set
• All actions fire within the randomised ranges we agree on and respect user-defined break keys
• No visible flicker, window focus change, or excessive CPU usage during a one-hour continuous run

I will share the target application and full workflow once we sign an NDA. If the brief fits your skill set, let’s discuss your approach and any libraries or design patterns you prefer to use.

Skills: Python, Software Architecture, Test Automation, C++ Programming, OpenCV

I have a Python-based AI video generator built with Streamlit, OpenCV, and FFmpeg.

Current issues:

unstable video pipeline (sometimes crashes)

audio and video are not always synced

problems with audio duration and merging

Scope:

debug and fix the existing pipeline (NOT building from scratch)

ensure stable video + audio output

improve reliability and prevent crashes

Requirements:

strong Python skills

experience with OpenCV

experience with FFmpeg (important)

experience with video processing pipelines

Nice to have:

worked on similar video automation systems

Goal:
Fix the current system and make it stable and usable.

We are seeking a skilled Computer Vision Engineer to develop a real-time object detection system using OpenCV and Python. The project involves integrating a camera module with a microcontroller to detect objects and send notifications. The ideal candidate will have experience in computer vision and embedded systems, with a strong understanding of object detection algorithms and microcontroller programming.

Looking for an experienced Android Java developer to build a complete app for an AR glasses device.

The device connects to an Android phone via USB-C. The app must:

— Open a live USB UVC camera feed fullscreen on the glasses display
— Apply real-time zoom from 1x to 10x with sharpness maintained
— Apply real-time contrast enhancement
— Apply real-time brightness adjustment
— Apply edge detection outline mode
— Apply colour filter modes
— Read visible text aloud using OCR and text to speech
— Simple on-screen menu to switch between features
— Maintain minimum 30 FPS at all times

SDKs provided:
— libuvc.aar — USB camera, zoom, image effects (Java API over C/C++ native libraries)
— GlassSDK.aar — glasses communication and display control (Java API over C/C++ native libraries)
— libuac.aar — USB audio output for text to speech (Java API over C/C++ native libraries)

All SDKs are Java based. Developer must be comfortable with Java Android development and JNI-based library integration.

Deliverables: working APK, GitHub link with clean code, screen recording showing all features.

Language: Java (Android) and C++

• Describe your recent experience with similar projects
• Include a link to your GitHub profile and/or website

We are seeking a highly capable Computer Vision / Document AI Engineer with strong AWS and banking / payments fraud experience to help design and build an intelligent check fraud detection solution.

This work is focused on modernizing fraud operations by replacing manual visual review of checks with a system that can identify anomalies, validate document integrity, and surface only high-risk exceptions for review. We are looking for someone who understands how to translate complex fraud requirements into a practical, production-oriented solution.

Scope of Work:

The selected consultant will help design and implement an AWS-native solution capable of processing check images, including front and back images, using OCR services such as Amazon Textract. The solution should be able to extract and validate key fields such as amount, account, date, payee, and endorsement-related data. It should also validate check structure, field placement, and document consistency, while detecting alterations, tampering, structural inconsistencies, and cross-field inconsistencies.

The role also requires comfort working on handwriting or signature analysis for suspicious endorsements, including comparison against known flagged fraud patterns where applicable. The solution should be designed with future integration in mind for external data sources such as account validation services, fraud flags, databases, and APIs.

Required Background:

We are interested in candidates with strong, hands-on experience across most of the following areas:

Computer Vision, Document AI, and OCR
Python, OpenCV, and PyTorch or TensorFlow
Amazon Textract and AWS services such as S3, Lambda, and SageMaker or similar
Banking, payments, check processing, or financial fraud detection
Handwriting or signature analysis, or comparable document verification use cases

What We Value:

We are looking for someone who has built production-oriented systems, not just experimental models or notebook-based prototypes. The strongest candidates will have experience working with real-world document and image quality issues, fraud detection workflows, rule-based validation combined with machine learning, financial document analysis, and enterprise API or backend integrations.

Equally important, we value clear thinking, strong execution, and the ability to work at a high technical standard in an environment where accuracy, reliability, and sound engineering judgment matter.

To Apply:

Please include your resume or CV, along with a brief summary of your relevant experience. We would also like to see examples of work involving OCR or document extraction, image anomaly detection, handwriting or signature analysis, AWS-based ML deployment, and fraud detection in banking or payments. Please also share your availability over the next few weeks.

OVERVIEW
We are building a web-based AI interior styling application with two core capabilities:
1. A template-driven layout system that arranges product images into a clean, structured visual composition
2. A photorealistic compositing system that places those same objects into real interior scenes with accurate lighting, scale, and spatial integration

Both systems must support prompt-based object replacement, allowing users to swap individual items dynamically without affecting the rest of the composition.

This project begins as a high-quality demo that will require approximately 30 hours of work and may expand into a full production application and pilot deployment.

NOTE: The budget is flexible for the right candidate.


CORE REQUIREMENTS
The highest level of visual realism is essential.

Final composited images must:
1. Match lighting direction, softness, and intensity
2. Include accurate shadows and grounding
3. Maintain correct scale and perspective
4. Integrate seamlessly into the environment with no artificial appearance

Outputs should meet the standard of professional interior photography, not typical AI-generated imagery.


SCOPE OF WORK
PHASE 1: Template-Based Layout System
1. Build structured templates that arrange product images into a clean visual composition (layouts will be provided)
2. Support individual object slots/regions
3. Enable prompt-based replacement:
4. Example: “replace chair with leather chair”

Ensure:
consistent spacing and alignment
proper scaling of each object
no overlap or distortion

PHASE 2: Photorealistic Compositing System
Take selected objects and place them into pre-mapped interior scenes

Work with:
predefined placement regions (walls, surfaces, floor zones, etc.)

Apply:
lighting matching
shadow generation
perspective correction
color and texture blending

Ensure outputs are indistinguishable from real photography

PHASE 3: Prompt-Based Replacement (Both Systems)
Enable dynamic updates where: individual objects can be swapped via text prompt, only the targeted region updates,
the rest of the composition remains unchanged


DEMO DELIVERABLES
1. Three structured layout templates:
design-focused
styling-focused
comparison-focused

2. Three fully composited interior images using mapped backgrounds

Functional prompt-based replacement in both systems


POST DEMO (Potential Expansion)
Full-stack application development
Scalable architecture for multiple users
Asset onboarding (large product catalogs)
Performance optimization for faster generation


REQUIRED SKILLS
1. Strong experience in image compositing and photorealistic rendering
2. Deep understanding of: lighting, shadows, and material interaction, scale and perspective in interior spaces
3. Experience with generative AI and/or image processing pipelines
4. Ability to build or contribute to structured, modular systems
5. Strong problem-solving and communication skills


PREFERRED EXPERIENCE
1. Computer vision (segmentation, masking, region detection)
2. Experience with interior or product visualization
3. Building interactive or user-driven visual tools
4. Frontend integration (React, WebGL, or similar)


TOOLS/TECHNOLOGIES (Examples)
Experience with some of the following is expected:
Python (preferred for image processing pipelines)
JavaScript / TypeScript
Image processing libraries (OpenCV, PIL, etc.)
Generative AI tools or APIs
Photoshop or similar (for prototyping or reference workflows)
Cloud platforms (AWS, GCP, or similar)


IMPORTANT NOTES
This is not a prompt-only project
This is a system design + visual realism problem
Precision and execution quality are more important than speed alone

I am looking for a Senior Computer Vision Engineer to build a high-performance engine for a production system. This is not a generic "find a cat or dog" project. The goal is to detect and track specific brand logos within professional event videos (sports, marathons, etc.) where the logo is often warped, angled, or on moving clothing.

The Task:
The system must take a Reference Logo (Image) and a Target Video as input. It must then return a detailed report including:

Timestamp Logs: Every second/frame the logo is visible.

Spatial Analytics: Precise placement/coordinates of the logo (e.g., on a helmet, a bib, or a background banner).

Robust Detection: The engine must handle non-rigid deformations (wrinkled clothes), extreme perspective shifts (side angles), and high-speed motion blur.

Technical Requirements:

Feature Matching Specialist: Must be an expert in Zero-Shot Detection and Descriptor Matching (e.g., YOLO-World, LightGlue, SuperPoint, or LoFTR). We cannot "re-train" the model for every new logo; it must work instantly upon upload.

Video Pipeline: Experience building efficient frame-processing pipelines (OpenCV, FFmpeg, CUDA).

Geometric Invariance: Proven ability to solve for perspective and affine transformations.

Tracking: Implementation of robust temporal tracking (DeepSORT, ByteTrack, or similar) to ensure "lock" during video pans and zooms.

Who I’m Looking For:
A developer who has built real-world production systems. You should understand how to optimize for accuracy (80%+) and handle the complexity of "messy" video environments.

How to Apply:
Show me a demo or describe a project where you successfully detected an object that was distorted, tilted, or moving at high speed.

I’m assembling a practical textbook that teaches drone technologies through hands-on programming with Python. The book should walk readers from fundamental concepts straight into real-time applications, letting them write, test, and fly code as they learn.

Core coverage has to include basic Python syntax for newcomers, real-time data processing techniques, and solid drone control algorithms that work on popular flight controllers (PX4, ArduPilot, DJI SDKs, etc.). Example projects—think live telemetry dashboards, autonomous waypoint missions, and computer-vision-assisted flight—should be coded step-by-step so readers can replicate results on their own hardware.

Because it’s a learning resource, clear explanations, commented code, troubleshooting tips, and small review exercises matter as much as the finished programs. Diagrams, screenshots, and short video demos (links are fine) will strengthen the lessons.

Please attach a detailed project proposal outlining:
• the chapter flow you suggest
• the real-time programming tools or libraries you plan to use (e.g., DroneKit-Python, MAVLink, PySerial, OpenCV)
• how you’ll validate that each script runs on actual hardware or a simulator

I’ll review proposals based on technical depth, clarity, and how engaging the learning experience will feel to a reader with a laptop, a drone, and a weekend to experiment.

Skills: Python, Linux, Troubleshooting, Technical Writing, Software Architecture, Software Development, OpenCV, Computer Vision

I need a rock-solid, real-time player tracking module for football matches that guarantees the ID assigned to each athlete at kick-off never changes until the final whistle. Right now, our OpenCV–TensorFlow–YOLO pipeline sometimes swaps or loses IDs when athletes overlap, leave the frame briefly, or the camera angle shifts, and that ruins every speed, distance, position, and heat-map metric we generate.

Key requirements
• Sport: football.
• Camera setup: five or more synchronized feeds.
• Existing stack: OpenCV, TensorFlow, YOLO – your solution must plug into this environment.

What I expect
1. A multi-object tracker with integrated re-identification that preserves the same unique ID through occlusion, crossings, short disappearances, or camera changes.
2. Cross-camera association so the same player is recognised seamlessly when switching angles.
3. Sub-second latency suitable for live analytics.
4. An admin panel where I can select a player, watch them in real time, and review historical metrics and footage.
5. Clear API hooks or Python modules so we can pipe the stable IDs into our existing performance-analysis dashboards.

Acceptance test
• 90-minute match recording with deliberate occlusions and camera switches: zero ID swaps permitted.
• CPU/GPU usage and frame rate documentation for typical 1080p streams on an RTX-class GPU.

If you have production experience with deep-SORT, ByteTrack, FairMOT, or custom Siamese re-ID networks and can fine-tune them for football dynamics, let’s talk. Provide a brief outline of your proposed approach, relevant past work, and timeline for a functional prototype.

Skills: C Programming, Python, Software Architecture, C++ Programming, OpenCV, Computer Vision, Deep Learning, YOLO

This is NOT a research project.
This is a real-world system where accuracy must approach 100%.

1. WHAT YOU WILL WORK ON
You will improve an existing automatically processing pipeline:
- Ball detection & tracking (different models; noisy, real-world conditions)
- Event detection:
Serve
Hit/Shots (contact ball-racket)
Bounce
Net
- Player tracking + pose estimation
- Temporal segmentation (play vs no-play)

2. CORE CHALLENGE
The problem IS:
“make the system 100% reliable in real conditions”

You must:
- Handle noise, occlusions, missing frames
- Use physics (ball trajectories)
- Use logic (tennis rules)
- Combine signals into a robust system

REQUIRED SKILLS
- Strong Python (OpenCV, NumPy)

Experience with:
- Object detection / tracking
- Multi-frame logic
- Ability to work with imperfect data
- Strong debugging skills

3. BIG PLUS
- Sports CV (tennis ideal)
- Ball tracking / trajectory modeling
- Pose estimation experience

4. IMPORTANT
This role is NOT for:
- Academic-only profiles
- People who only trained models on clean datasets

This role IS for:
Engineers who can make systems work in reality.

5. APPLICATION PROCESS
If you apply, you will receive a technical test (labeling and/or Ground Truth cleaning).
No long interviews.
Your work will speak.

6. PROJECT STRUCTURE
Start: small paid task (1–2 days)
If strong → long-term collaboration

7. TO APPLY
Send:
- Short intro (max 5 lines)
- 1 project where you improved a system (not just built)
- GitHub (if available)

8. FINAL NOTE
If you are looking for:
- Clean datasets
- Academic experiments

This is NOT for you!

If you want to:
build a real system that works under pressure
Apply.

We're a product company building an AI-powered operations platform. We need an experienced engineer to help us design and implement an end-to-end AI pipeline that combines visual data processing with automated decision-making

What We're Building:
- A system that ingests image/video input, extracts structured data, and triggers automated workflows
- AI-assisted categorization, enrichment, and output formatting
- Human-in-the-loop review layer before final actions are taken
- Scalable, modular architecture that can grow with the product

Who We're Looking For:
- Experience building autonomous AI agents or automated pipelines (not just chatbots)
- Comfortable with computer vision tools (OpenCV, YOLO, or similar)
- Has worked with LLMs (Claude, GPT-4, OpenAI API) in production
- API integrations and workflow orchestration
- Can work independently from a roadmap and document cleanly

Nice to Have:
- Experience with image generation or style consistency (Midjourney, Stable Diffusion)
- Background in eCommerce or ops-heavy businesses
- Supabase or vector databases for knowledge base

To Apply: share a relevant project, what the system did, what tools you used, and the outcome. Loom or case study welcome

I have a set of GelSight images and need an end-to-end workflow that turns them into accurate, high-resolution 3D meshes of everyday objects. The goal is to capture subtle surface variations—scratches, fine textures, and edge curvature—so the final model can be inspected, measured, and exported for downstream use in CAD or visualization tools.

You may choose whichever environment you are most comfortable with (Python + OpenCV/PyTorch, MATLAB, or C++). What matters is that the pipeline:

• takes raw GelSight data as input (I will supply sample files),
• performs all necessary photometric calibration and image alignment,
• reconstructs a dense height map, then converts it to a watertight mesh,
• outputs OBJ or STL alongside a brief report on reconstruction quality.

Results should preserve features down to roughly 10 µm where the source data supports it; I will verify by comparing cross-sections against a reference profile.

Please include an outline of your planned approach, example of a similar reconstruction you have completed, and the estimated turnaround time for a first working prototype.

Skills: Python, Matlab and Mathematica, 3D Rendering, C++ Programming, 3D Modelling, 3D Design, Computer Vision, 3D CAD

I need a clear, evidence-based report that compares the performance of today’s most widely cited object-detection algorithms. The focus is strictly on Computer Vision, zeroing in on Object Detection, and the core goal is to evaluate how the main approaches stack up against each other in terms of accuracy, speed, computational cost, and real-world suitability.

Scope
• Analyse at least three state-of-the-art methods—think Faster R-CNN, SSD, YOLO (v7/8), DETR or similar.
• Draw all claims from peer-reviewed journals, top-tier conference papers, or authoritative benchmark leaderboards (e.g., COCO, PASCAL VOC).
• Present metrics consistently (mAP, FPS, FLOPs, params, latency) so direct comparison is effortless.
• Highlight strengths, weaknesses, and trade-offs instead of simply listing numbers.

Writing & Formatting
The narrative must read like a human-authored technical paper, without detectable AI-generated phrasing. Use a neutral, objective tone, incorporate figures or tables where they genuinely aid clarity, and cite all sources meticulously. I accept either APA 7th or IEEE style—pick one and stay consistent.

Deliverables
1. A polished 3 000–5 000-word report in editable Word (or LaTeX PDF) plus a clean PDF.
2. All reference entries and in-text citations formatted correctly.
3. A short appendix detailing any datasets, code repos, or tools referenced (TensorFlow, PyTorch, OpenCV, etc.).
4. Turnitin similarity report and confirmation that the text passes common AI-content detectors (GPTZero, Originality.ai).

Acceptance Criteria
• Every statistic is traceable to a cited, peer-reviewed source.
• Comparative discussion is data-driven, not anecdotal.
• Language flows naturally and survives AI-detection checks.
• Final document is free of grammatical errors and formatting glitches.

If this sounds like the kind of deep-dive you enjoy crafting, let’s get started—I’m ready to review your outline and timeline.

Skills: Research, Technical Writing, Ghostwriting, Research Writing, Machine Learning (ML), Data Analysis, Computer Vision, Object Detection

I need a clear, evidence-based report that compares the performance of today’s most widely cited object-detection algorithms. The focus is strictly on Computer Vision, zeroing in on Object Detection, and the core goal is to evaluate how the main approaches stack up against each other in terms of accuracy, speed, computational cost, and real-world suitability.

Scope
• Analyse at least three state-of-the-art methods—think Faster R-CNN, SSD, YOLO (v7/8), DETR or similar.
• Draw all claims from peer-reviewed journals, top-tier conference papers, or authoritative benchmark leaderboards (e.g., COCO, PASCAL VOC).
• Present metrics consistently (mAP, FPS, FLOPs, params, latency) so direct comparison is effortless.
• Highlight strengths, weaknesses, and trade-offs instead of simply listing numbers.

Writing & Formatting
The narrative must read like a human-authored technical paper, without detectable AI-generated phrasing. Use a neutral, objective tone, incorporate figures or tables where they genuinely aid clarity, and cite all sources meticulously. I accept either APA 7th or IEEE style—pick one and stay consistent.

Deliverables
1. A polished 3 000–5 000-word report in editable Word (or LaTeX PDF) plus a clean PDF.
2. All reference entries and in-text citations formatted correctly.
3. A short appendix detailing any datasets, code repos, or tools referenced (TensorFlow, PyTorch, OpenCV, etc.).
4. Turnitin similarity report and confirmation that the text passes common AI-content detectors (GPTZero, Originality.ai).

Acceptance Criteria
• Every statistic is traceable to a cited, peer-reviewed source.
• Comparative discussion is data-driven, not anecdotal.
• Language flows naturally and survives AI-detection checks.
• Final document is free of grammatical errors and formatting glitches.

If this sounds like the kind of deep-dive you enjoy crafting, let’s get started—I’m ready to review your outline and timeline.

Skills: Research, Technical Writing, Ghostwriting, Research Writing, Machine Learning (ML), Data Analysis, Computer Vision, Object Detection

I need a compact, well-structured script that ingests a 2D image and returns a high-quality .glb file in one pass, with no noticeable loss of fidelity. Speed is critical; the final tool should feel almost instantaneous when run on a modern desktop.

Because I have no fixed preference for specific libraries, feel free to choose whatever stack best balances performance and quality—OpenCV, TensorFlow, or any other framework you’re confident will get the job done quickly and cleanly. What matters most is that the resulting geometry, textures, and materials in the .glb look crisp and faithful to the source image.

Deliverables
• Source code with clear inline comments
• A straightforward README explaining environment setup, usage, and any dependencies
• A sample run demonstrating the conversion of at least one provided 2D image into a .glb, including screenshots of input and output
• Brief performance metrics (time taken and file size) on a typical image so I can benchmark it on my machine

Acceptance Criteria
• Conversion completes in under 60 seconds on a mid-range CPU/GPU combo
• Output .glb opens without errors in standard viewers (e.g., Blender, Windows 3D Viewer)
• Textures remain sharp; no visible artifacts or compression noise

If you can deliver a blazing-fast, high-fidelity pipeline with concise code and solid documentation, let’s get started.

Skills: C Programming, Python, CUDA, C++ Programming, 3D Modelling, OpenCV, 2D Drawing, AI Development

I want to build a object size measurement application that utilizes structured light depth or stereo camera to measure sizes of different objects at close distance (less than 1.5m) with high accuracy (mm level) and high performance.

Proposals should include required timeline, the most efficient approach to achieve this and previous experience with depth analysis and size estimations.

• Describe your recent experience with similar projects
• Include a link to your GitHub profile and/or website
• What frameworks have you worked with?

We are looking for a specialized Deep Learning Engineer to develop a robust pipeline for the automated analysis of Brain MRI scans. The primary goal is to perform two main tasks:

Semantic Segmentation: Precisely masking Glioma regions (Tumor Core, Edema, Enhancing Tumor).

Grade Classification: Categorizing the detected tumors into Low-Grade Glioma (LGG) or High-Grade Glioma (HGG).

Technical Requirements & Scope

Architectures: Implementation of U-Net (or variants like Attention U-Net/V-Net) for pixel-wise segmentation and EfficientNetB3 (or similar CNNs/ViTs) for classification.

Data Handling: Processing multi-modal MRI sequences (T1, T1ce, T2, and FLAIR) using libraries like Nibabel, SimpleITK, or OpenCV.

Preprocessing: Experience with N4 Bias Field Correction, intensity normalization, and data augmentation tailored for medical imaging.

Frameworks: Proficiency in PyTorch or TensorFlow/Keras.

Evaluation: Focus on metrics such as Dice Coefficient, IoU for segmentation, and Accuracy/F1-score for classification.

Deliverables

A trained model (inference-ready).

Preprocessing and post-processing scripts.

Documentation of the model performance and evaluation metrics.

**Project Title:**
Precision Extraction of Black Overprints from High-Resolution Stamp Sheets (Research Project)

**Project Description:**
I am conducting a specialized philatelic research study on 1931 Zeppelin overprints. I have two full stamp sheets (50 stamps each, high-resolution scans provided as PDF files), and I need the overprint isolated from each individual stamp with maximum fidelity. This is NOT a graphic design or basic background removal job. This is a precision image extraction task for research purposes requiring strict adherence to instructions and preservation of original detail.

The source files are high-resolution PDF scans. You must extract images at native resolution (e.g., 1200 dpi). Screenshots, resampling, or any quality loss are not acceptable. If needed, convert the PDFs into high-resolution TIFF or PNG files prior to processing.

Scope of work includes processing 2 full sheets (50 stamps per sheet), segmenting each sheet into individual stamps (total 100 stamps), extracting ONLY the overprint from each stamp, and delivering each overprint as a separate image file. The provided 6 PDF files represent segments of the full sheets and must be digitally reconstructed into complete sheets before assigning positions.

CRITICAL REQUIREMENTS: The overprint is in black ink and is often broken, faint, or incomplete. The scans include red pointer arrows that must be removed. Under no circumstances should the overprint be reconstructed, enhanced, redrawn, cleaned, or visually improved. Do not fill in missing parts of letters or numbers. Do not smooth, sharpen, or alter edges. All natural defects, breaks, and irregularities must be preserved exactly as they appear in the original scan. Any work where the overprint appears artificially completed, repaired, or overly clean will be rejected.

Technical expectations include use of masking, color separation, or channel-based extraction techniques. Preserve fine detail and edge irregularities. Avoid aggressive thresholding that removes weak ink and avoid noise reduction that removes real features. A reproducible workflow (script, macro, or clearly documented method) is preferred.

File Naming & Position Mapping Requirements: Each stamp must be labeled using the format S[Sheet]_R[Row]_C[Column].png. Example: S1_R03_C04.png corresponds to Sheet 1, Row 3, Column 4. Each sheet must be structured as a 10 × 5 grid, with rows numbered top to bottom (R01–R10) and columns left to right (C01–C05). The original sheet orientation must be preserved exactly; do not rotate, flip, or reorient images. Row 1 must correspond to the top row of the original scan. The freelancer must reconstruct the full sheets from the provided PDF segments before assigning row and column positions. Deliver files organized into folders by sheet (e.g., /Sheet1/ and /Sheet2/).

**Visual Grid Overlay Requirement:**
Before final extraction, the freelancer must create and share a visual grid overlay on each reconstructed full sheet showing the 10 × 5 stamp layout and assigned positions. Each stamp position must be clearly labeled in the format S1_R01_C01 through S1_R10_C05 for Sheet 1 and S2_R01_C01 through S2_R10_C05 for Sheet 2. Rows must run from top to bottom and columns from left to right, based strictly on the original scan orientation. This overlay is for verification before final delivery and is intended to confirm that sheet reconstruction and position mapping are correct. The overlay may be provided as a JPG, PNG, or PDF preview. Final extraction should proceed only after the position map is clearly established. I prefer that each reconstructed sheet first be delivered as a labeled preview image with all 50 positions marked, so I can confirm row/column mapping before the 100 final extracted files are produced.

Deliverables include 100 individual image files (one per stamp overprint), preferably as PNG files with transparent background or clean black/white output, with correct positional naming. Each extracted overprint must maintain consistent scale and alignment relative to the original stamp; do not resize, distort, or inconsistently crop individual outputs. Output background must be either fully transparent or a clean uniform background (pure white or pure black), with no residual artifacts from the original stamp design. Output resolution must match the original extracted stamp resolution, with no downscaling or compression. Each output must be visually reviewed to ensure no loss of faint ink and no accidental removal of valid overprint elements.

Final delivery must be provided as a structured ZIP archive containing Sheet1 and Sheet2 directories, along with a CSV or Excel file mapping each filename to its sheet, row, and column position. Optional deliverables include layered source files (PSD or equivalent) and a brief explanation of the workflow used. Fully automated batch processing without manual verification is not acceptable; each output must be individually checked for accuracy.

Selection process: Shortlisted candidates will be required to extract one sample overprint from a provided image before the full project is awarded. This is mandatory to verify accuracy and adherence to requirements.

Ownership & Confidentiality: All source images and resulting outputs are confidential and remain the exclusive property of the project owner. The freelancer may not reuse, distribute, or publish any materials or results from this project without permission.

To apply, please include examples of similar image extraction or segmentation work, the tools/software you plan to use, and your estimated turnaround time.

Skills: Python, GIMP, Image Processing, OpenCV, Computer Vision, Adobe Photoshop

We are hiring a professional development team (NOT individuals) to build a complete production-grade platform for structured content generation.

This is a serious, fully scoped project with strict technical requirements and a fixed budget.



📌 PROJECT OVERVIEW

We are building a unified system that transforms structured narrative input into:

• Books (structured, canonical data)
• Multi-character audiobooks (voice + timing)
• Cinematic motion-novel episodes (visual rendering)
• Promotional content (clips, trailers, assets)

This is NOT a simple app.

This is a deterministic pipeline system where:
Input → Processing → Output is fully controlled, reproducible, and validated.


⚠️ IMPORTANT — READ BEFORE APPLYING

This system:

• MUST be deterministic (same input = same output)
• MUST NOT rely on generative AI randomness
• MUST NOT use OpenAI or similar APIs for core logic
• MUST NOT generate content unpredictably

We are building a CONTROLLED pipeline.


SYSTEM STRUCTURE

The platform includes 4 integrated modules:

1. Book Engine
2. Audio Engine
3. Visual Engine
4. Promotion Engine

All modules:

• Work together in one platform
• Share structured data (IDs, schema)
• Must also function independently
• Must export outputs at every stage


PIPELINE FLOW

The system MUST follow this exact flow:

1. Full book creation (story, characters, structure)
2. Full-book voice assignment (before any audio)
3. Audio generation per chapter
4. Full-book visual validation (characters, versions, assets)
5. Visual episode generation (synced to audio)
6. Promotion content generation (clips, trailers, posters)

NO skipping steps.
NO partial pipeline execution.


🎬 VISUAL RULES

• NO dynamic generation of visuals
• Only predefined assets allowed
• Character consistency MUST be enforced
• Templates MUST be used (15–20 minimum)

🎧 AUDIO RULES

• All audio allowed (no censorship)
• Voice assigned ONCE per character (full book)
• 10 primary voices + secondary voice system
• Emotion-driven non-verbal system (deterministic)


📢 PROMOTION SYSTEM

Must generate:

• Clips
• Trailers
• Posters

Rules:

• NO content generation (only transformation)
• Must preserve meaning
• Must include subtitle sync (≤100ms)

Music:

• Background music allowed
• Dialogue must always remain clear (auto-ducking required)

TECH STACK

You can use:

• FFmpeg
• OpenCV
• MoviePy
• Coqui / XTTS / RVC
• Whisper / Aeneas
• Librosa / PyDub

You are NOT allowed to build core engines from scratch.


TEAM REQUIREMENT

We ONLY work with a full team:

• 1 Systems Architect
• 3 Backend Engineers
• 2 Frontend Engineers
• 1 Media Engineer
• 1 QA Engineer
• 1 DevOps Engineer

Total: 9 people

Rules:

• All members must be real (no fake team)
• Each must be available for direct communication
• Each must attend meetings when required
• Each must sign NDA + agreement
• No outsourcing or hidden contributors

PROJECT DETAILS

Timeline: Maximum 10 weeks
Budget: $7,500 (development)

Additional:
• $500 for voice actors (handled by us)


ACCEPTANCE REQUIREMENTS

System must:

• Run full pipeline end-to-end
• Process 30+ chapters
• Be fully deterministic
• Have ZERO missing features
• Require NO manual fixes

Anything less = rejected.


WHAT WE DO NOT ACCEPT

• Partial systems
• Mock/demo builds
• Hidden scripts/manual fixes
• AI-based randomness
• “We will fix later” delivery
• Individual freelancers pretending to be a team


HOW TO APPLY

include:

1. Your FULL team structure
2. Relevant experience (pipelines, media processing)
3. Tech stack confirmation
4. Confirmation you can meet timeline & budget


FINAL NOTE

This is a strict, high-standard project.

If you are not experienced with:
• system architecture
• media pipelines
• deterministic processing

→ Do NOT apply.

We are looking for a team that can execute at production level.

We are looking for a person to help embrace AI in our Development department, to be a motivator and show practical examples and help automate the complete dev cycle using langgraph

We are building a structured content production platform and are looking for a small experienced developers or team to build a Visual Engine.

This system converts structured inputs into:

- cinematic motion-novel episodes
- reusable visual assets
- production-ready video outputs

This is not an AI video generation system.
It is a deterministic rendering and composition pipeline that must strictly follow structured inputs and approved assets.

Scope of Work:

The system must include:

- Character and location registry system
- Pre-production extraction and approval workflow
- Template-based shot composition system
- Visual plan system (scene → shot mapping)
- Timeline synchronization with audio
- Rendering and episode assembly pipeline
- QC and validation system
- Export system (video + structured outputs)
- Manifest generation
- Full UI for non-technical operators


Technical Requirements:

- Strong experience with FFmpeg
- Experience with OpenCV and MoviePy
- Experience with PIL / Pillow and NumPy
- Experience building rendering or video pipelines
- Strong understanding of structured data systems

The system must:

- NOT generate narrative or creative content
- strictly follow structured input data
- operate deterministically (same input = same output)
- ensure perfect timing sync with audio

Important System Requirement:

Every exported video must:

- automatically include the platform intro at the beginning
- ensure correct placement and synchronization
- fail export if intro is missing or incorrect


Project Structure:

- Fixed-price project
- Full detailed technical scope document will be provided
- System must operate fully standalone
- Must be UI-based (no CLI or manual backend work)


Team Requirement (MANDATORY):

Minimum required team:

- Rendering Engineer
- Backend Engineer
- Frontend Engineer
- QA Engineer

All roles must be actively assigned and available throughout the project.


Timeline:

- Target: 6–8 weeks
- Maximum: 10 weeks

Continuous development progress is required.


Budget:

- Fixed price: $2,000
- No upfront payment
- Full payment only after complete delivery and approval


Before Applying, You Must Confirm:

Please answer the following in your proposal:

1. Can you deliver within 6–8 weeks (max 10 weeks)?
2. Can you work within a fixed budget of $2,000?
3. Do you have the full required team (Rendering, Backend, Frontend, QA)?
4. Do you have experience with video pipelines, FFmpeg, OpenCV, or MoviePy?
5. Can you build a fully deterministic system (no randomness or AI generation)?
6. Can you deliver a complete system only (no partial delivery)?


A full technical scope document is already prepared and will be shared with shortlisted candidates.

Only apply if you can meet the full scope, team requirements, and delivery expectations.

I'm looking for an experienced programmer to help me set up a Raspberry Pi personal assistant. This project involves coding and configuring the Raspberry Pi, which is equipped with a camera and microphone, similar to Loona the pet robot.

Key functionalities include:
- Facial Recognition: Identify and recognize individuals.
- Object Detection: Recognize and categorize different objects in the environment.

Ideal skills and experience:
- Proficiency in Python and Raspberry Pi
- Experience with OpenCV or similar libraries for facial recognition and object detection
- Familiarity with setting up and configuring Raspberry Pi for various projects
- Strong problem-solving skills and attention to detail

Please provide examples of similar work done and ensure you can meet the requirements.

Skills: Python, Linux, Matlab and Mathematica, Arduino, Raspberry Pi, OpenCV, Facial Recognition, Object Detection

Hi! I'm building a tool called PlasmaEdge that uses AI to automatically find the best recipe (gas ratios, pressure, power settings) for plasma etching — a process used to carve nanoscale patterns into materials like silicon nitride and gallium nitride.

Right now, engineers often run hundreds of trial-and-error experiments to dial in the right settings. PlasmaEdge reduces that to ~10–15 intelligently guided runs.

Here’s how it works:

The system collects real-time process data from plasma chamber sensors (pressure, RF power, gas flow rates, temperature)
It applies eigenvalue analysis (PCA) to identify which variables actually drive etch performance — reducing 10+ parameters to the 2–3 that matter most
After each run, SEM images of the etched structures are analyzed to extract quantitative metrics such as etch rate, sidewall angle, edge uniformity, and surface roughness
A Bayesian optimization engine (Gaussian processes) uses that feedback to select the next best recipe — balancing exploration vs. exploitation
The loop repeats until optimal conditions are reached

Example outcome: As shown in the SEM images above, different process recipes produce dramatically different etch profiles — from rough, tapered, and inconsistent sidewalls to highly vertical, uniform, and clean structures. PlasmaEdge directly learns from these image-based differences to converge on recipes that yield the highest-quality etch results.

What I’m looking to build:
Real-time data pipeline for sensor ingestion and organization
PCA / eigenvalue module for dimensionality reduction
Bayesian optimization engine (Gaussian processes)
SEM image analysis pipeline for extracting etch quality metrics
Simple dashboard to visualize optimization progress
Skills needed:
Python (scikit-learn, BoTorch or GPyOpt)
PCA / eigenvalue decomposition
Bayesian optimization + Gaussian processes
Data pipeline development
Image processing (OpenCV or similar)
Bonus: semiconductor / manufacturing experience
Project details:
9-month federally funded R&D project
Part-time or full-time
Must be US-based
Compensation: negotiable

If you love applying AI to real-world engineering problems, I’d love to connect — please share relevant experience or similar projects you’ve worked on.

I am designing a system to develop a scanner using a Lidar sensor. I am looking for an expert in computer vision, machine learning, OpenCV, and data processing to assist with this project. The ideal candidate will have experience with Lidar technology and be able to integrate it into a comprehensive system. Responsibilities include developing algorithms for data processing and ensuring the system's accuracy and efficiency.

We are looking to hire experienced freelancers to help build a next-generation AI-powered camera management system for a camera business that is moving from traditional operations to a more advanced, centralized, and intelligent platform.

The objective is to develop a scalable solution that can support multi-camera monitoring, AI-based body scanning, behavior tracking, vehicle number plate recognition, drone integrations, real-time alerts, and secure admin control. The platform should be designed for future growth and long-term operational use. This is not a basic prototype — we are looking for professionals who have worked on relevant, production-level systems.

Project Scope
The system will include, but is not limited to, the following capabilities:
1. Live camera streaming and recording.
2. AI-based body/person scanning and detection.
3. Behavior tracking and suspicious activity analysis.
4. Number plate detection and recognition.
5. Drone integration for extended surveillance coverage.
6. Smart event detection and automated notifications.
7. Admin dashboard with user, camera, and access management.
8. Secure data handling, audit logging, and role-based access.
9. Scalable architecture for future expansion and advanced features.

Required Expertise:

We are specifically looking for freelancers with hands-on experience in one or more of the following areas:
-Backend development for surveillance, monitoring, or AI-based systems.
-Frontend development for admin dashboards and monitoring interfaces.
-Mobile app development for iOS and Android.
-AI / ML development for detection, recognition, and tracking.
-Computer vision and video analytics.
-Camera integrations using RTSP / ONVIF.
-Drone integration and external device connectivity.
-Vehicle number plate recognition / LPR systems.
-Cloud architecture, DevOps, and deployment pipelines.
-QA and testing for real-time and AI-driven systems.

Preferred Technical Stack Experience

Experience with the following technologies will be highly valuable:
-React.js or similar dashboard frameworks.
-Node.js, Python, or similar backend stacks.
-OpenCV, TensorFlow, PyTorch, YOLO, FaceNet, DeepFace, EasyOCR, or OpenALPR.
-FFmpeg, WebRTC, or GStreamer for video processing.
-Docker, Kubernetes, CI/CD, and cloud deployment.
-Secure authentication, RBAC, logging, and scalable system architecture.

Who Should Apply:
We are looking for freelancers who can contribute to a serious, real-world, production-grade platform. This project requires people who understand the complexity of camera systems, AI analytics, video processing, and integration-heavy environments.

Important Note:
Only freelancers who can share directly relevant project experience will be considered. Please apply only if you have completed work that closely matches this requirement. Profiles without matching experience or unrelated examples will be ignored.

How to Apply:
-Please include the following in your response:
-A brief summary of your relevant experience.
-Details of similar projects you have completed.
-Links to live projects, portfolios, demos, or case studies.

Skills: PHP, Mobile App Development, Android, Software Architecture, Machine Learning (ML), Backend Development, Security Camera, Computer Vision, AI Development

Join our team as a Senior Video Streaming Engineer to enhance our football analytics platform. You'll work with video data, ensuring seamless streaming and processing for real-time analysis. Collaborate with our team to integrate video feeds into our system, focusing on performance and reliability. This role requires expertise in video streaming technologies and a passion for sports analytics.

I need a computer-vision pipeline that can ingest an uploaded logo image (PNG or JPG) together with a video file and then return every instance of that logo throughout the clip. The videos that come in will be MP4, AVI, or MOV, and the solution must run smoothly.

Here’s how I picture the workflow:
• The script accepts the logo and video paths from a simple CLI or minimal API endpoint.
• It scans each frame, flags every logo hit, and outputs a tidy JSON or CSV that lists, for each detection, the start timestamp, on-screen duration, and bounding-box coordinates.
• No extra training or fine-tuning whenever a new logo is supplied—the detector should work right away with whatever image is uploaded.

I’m after well-commented, maintainable code, and I’ll need brief setup instructions so I can reproduce your results locally (Python virtual-env, requirements.txt, maybe a Dockerfile). Fast turnaround is important—I’d like to start reviewing working code as soon as possible.

If you’ve already built anything similar—logo finders, object matchers, ORB/SIFT/SURF pipelines, YOLO or Faster-RCNN tweaks—tell me about it. Strong communication during the build is a must, and please start your reply with “SPORTS” so I know the details above were read.

Skills: Python, Linux, Software Architecture, Machine Learning (ML), Ubuntu, OpenCV, Computer Vision, Deep Learning, Object Detection, YOLO

We are building a short-range detection system for small UAVs / drones at approximately 110 meters.

Radar will be the primary sensor. A thermal / infrared camera will be used for target confirmation, false alarm reduction, and improved tracking accuracy.

We are looking for a contractor who can implement a vendor-agnostic MVP fusion algorithm that works on recorded radar data and synchronized thermal video in real time. The exact radar model has not been finalized yet, so the first phase must not depend on one specific radar vendor or SDK.

Project scope:
- ingest and normalize radar detections / tracks and thermal video or frame data;
- synchronize radar and camera data in real time;
- use existing or estimated radar-to-camera alignment / calibration;
- implement camera-assisted confirmation of radar detections;
- implement fused tracking and confidence scoring;
- provide readable code, configs, README, and a short technical report.

Preferred experience:
- sensor fusion;
- radar perception or FMCW radar data processing;
- thermal / infrared computer vision;
- multi-object tracking;
- camera calibration and radar-to-camera alignment;
- Python, OpenCV, NumPy, SciPy;

Expected deliverables:
- source code;
- reproducible pipeline;
- config files;
- sample run on our data;
- output files with fused tracks;
- demo video with overlays;
- short report explaining the method, assumptions, limitations, and next steps.

Out of scope for this MVP:
- embedded deployment;
- production integration with a specific radar SDK;

Please include in your proposal:
1. A brief description of similar work you have done.
2. Your suggested MVP pipeline for radar + thermal fusion.
3. Your approach to time synchronization and calibration.
4. A milestone breakdown.
5. Relevant code samples if you have, GitHub, demo videos, or publications, if available.

We are seeking a skilled developer or small team to create a private API tool that takes any image (JPEG/PNG/SVG/PDF etc..) logo as input and returns two optimized versions of that logo:

- One version that looks visually clean and sharp on a white shirt.
- One version that works well on a black shirt.

This project involves image processing and potentially AI-based color adaptation to ensure logo visibility and aesthetic consistency across both backgrounds. The API must process the logo intelligently — without requiring manual intervention.

Key Features:

- RESTful API endpoint that accepts any image format.
- Returns two PNGs (white-shirt version, black-shirt version).
- Automatic detection of logo visibility issues (e.g. dark logos on black).
- Smart adjustments: remove background + enhance + delete free space + adoption to both backgrounds.

Lightweight and scalable architecture (hosted on AWS, Vercel, etc. optional).

Ideal Candidate Should Have Experience In:

- Image Processing Libraries: OpenCV, PIL, or scikit-image.
- Python or Node.js for backend/API development.
- Machine Learning / AI: (optional) for advanced color adaptation or enhancement.
- Color Theory & Contrast Algorithms (e.g., WCAG contrast compliance).
- Handling alpha/transparency.

Working with cloud functions / serverless APIs (e.g., AWS Lambda, Google Cloud Functions).

Deliverables:
- Fully functional API endpoint.
- Clear documentation on usage.

Bonus if You Have:
- Experience with logo detection or vector tracing.

To Apply:
Please share:

- Relevant experience or similar projects.
- Your proposed tech stack.

Estimated timeline and budget.

Any initial thoughts on how you'd approach contrast optimization.

• Have you worked with any computer vision models?
• what's your computer vision experience?