AI Computer Vision Analysis MVP Development
Budget: $15 – $25 USD
Summary
We are building a tightly scoped computer vision MVP for an internal product. This project is focused on analyzing pre-recorded workout videos to demonstrate core technical capability for identifying people, classifying exercises, and counting repetitions using pose estimation.
This is an applied engineering project, not academic research and not a production deployment. The initial output will be used internally for demonstration and board review.
What you will work on:
You will build a Python-based pipeline that processes workout video files we provide and outputs structured analysis results. Specifically, the system should:
• Identify known individuals in video clips using facial embeddings (closed-set recognition)
• Classify the exercise being performed from a small, fixed set of exercises
• Count repetitions for at least one exercise (for example squats) using pose-based heuristics
• Output results in a structured format including identified user, exercise label, rep count, per-rep timestamps, and confidence scores
A simple script or lightweight demo output is sufficient. UI polish is not a focus.
Scope constraints:
• Pre-recorded video only (no live streaming)
• Single primary subject per clip
• Small known user set (we will provide enrollment images)
• Limited exercise set (approximately 5–8 exercises)
• Rep counting required for at least one exercise
• Python implementation
• Cloud or local GPU friendly (mobile or on-device optimization is out of scope)
Deliverables:
• Python code implementing the full video analysis pipeline
• Clear separation between face identification, exercise classification, and rep counting logic
• Sample outputs on provided video files
• Short README explaining how to run the pipeline and describing the architecture
Required experience:
• Strong Python skills
• Computer vision fundamentals
• Experience with video processing and OpenCV
• Experience with pose estimation
• Experience with deep learning frameworks such as PyTorch or TensorFlow
• Ability to design clean, modular systems
Nice to have:
• Facial recognition or embedding-based identity systems
• Human activity or action recognition experience
• Prior work in fitness, sports, or biomechanics
• GPU inference or model optimization experience
Engagement details:
• Start on Monday (February 23)
• Duration approximately 4–6 weeks
• Hourly contract
Briefly describe how you would approach pose-based repetition counting for a squat using video.
When applying, please include links to relevant computer vision or machine learning projects you have worked on.
We are building a tightly scoped computer vision MVP for an internal product. This project is focused on analyzing pre-recorded workout videos to demonstrate core technical capability for identifying people, classifying exercises, and counting repetitions using pose estimation.
This is an applied engineering project, not academic research and not a production deployment. The initial output will be used internally for demonstration and board review.
What you will work on:
You will build a Python-based pipeline that processes workout video files we provide and outputs structured analysis results. Specifically, the system should:
• Identify known individuals in video clips using facial embeddings (closed-set recognition)
• Classify the exercise being performed from a small, fixed set of exercises
• Count repetitions for at least one exercise (for example squats) using pose-based heuristics
• Output results in a structured format including identified user, exercise label, rep count, per-rep timestamps, and confidence scores
A simple script or lightweight demo output is sufficient. UI polish is not a focus.
Scope constraints:
• Pre-recorded video only (no live streaming)
• Single primary subject per clip
• Small known user set (we will provide enrollment images)
• Limited exercise set (approximately 5–8 exercises)
• Rep counting required for at least one exercise
• Python implementation
• Cloud or local GPU friendly (mobile or on-device optimization is out of scope)
Deliverables:
• Python code implementing the full video analysis pipeline
• Clear separation between face identification, exercise classification, and rep counting logic
• Sample outputs on provided video files
• Short README explaining how to run the pipeline and describing the architecture
Required experience:
• Strong Python skills
• Computer vision fundamentals
• Experience with video processing and OpenCV
• Experience with pose estimation
• Experience with deep learning frameworks such as PyTorch or TensorFlow
• Ability to design clean, modular systems
Nice to have:
• Facial recognition or embedding-based identity systems
• Human activity or action recognition experience
• Prior work in fitness, sports, or biomechanics
• GPU inference or model optimization experience
Engagement details:
• Start on Monday (February 23)
• Duration approximately 4–6 weeks
• Hourly contract
Briefly describe how you would approach pose-based repetition counting for a squat using video.
When applying, please include links to relevant computer vision or machine learning projects you have worked on.