BenchMetric — Bench Press Analyzer (Streamlit + MediaPipe)

BenchMetric is a small computer-vision app that analyzes bench press technique from an uploaded video. It uses MediaPipe Pose for landmark detection and OpenCV for frame processing, then produces per-rep metrics and charts in a Streamlit UI.

Features

• Upload .mp4 bench press videos in the browser (Streamlit)
• Frame-by-frame analysis (OpenCV + MediaPipe)
• Phase detection: START / ECCENTRIC / BOTTOM / CONCENTRIC
• Per-rep/per-phase metrics:
  • Duration_s, Velocity_avg, Distance_px
  • Bottom_pause, Asymmetry
• Visual output:
  • live preview while analyzing
  • metrics table
  • velocity / duration charts
• Saves results to results.csv

Requirements

• Python 3.10+ (Python 3.11 recommended)
• Dependencies from requirements.txt

Note: this repo uses opencv-python-headless (great for Docker and servers without a GUI). If you want to run the non-Streamlit mode with cv2.imshow, you’ll typically want opencv-python instead.

Quickstart (local)

Install dependencies:

pip install -r requirements.txt

Run the Streamlit app:

streamlit run app.py

Then open http://localhost:8501, upload a video, and click Analyze video.

Run with Docker

Build the image:

docker build --no-cache -t bench-metric .

Run the container:

docker run --rm -p 8501:8501 bench-metric

Open http://localhost:8501.

How to use

1. Open the app.
2. Upload an .mp4 file.
3. Adjust settings:
   • Delay — artificial slowdown (useful for watching the analysis).
   • Right mode — analyze the right side (default is left).
4. Click Analyze video.
5. After it finishes, you’ll see:
   • a DataFrame with metrics
   • charts
   • results.csv written in the project directory

Output (results.csv)

The CSV is generated automatically after analysis. Columns include:

• Repetition — repetition index
• Phase — movement phase
• Duration_s — phase duration
• Velocity_avg — average velocity in the phase (px/s)
• Distance_px — bar path distance in the phase (px)
• Bottom_pause — bottom pause detected (0/1)
• Asymmetry — wrist asymmetry detected (0/1)

Project structure

• app.py — Streamlit UI (upload, progress, charts)
• main.py — video processing loop (OpenCV) + progress reporting
• pose_module.py — pose logic, phase detection, metrics, CSV export
• assets/ — sample images and videos

Troubleshooting

Progress bar is stuck / very laggy

• Video analysis is CPU-bound and runs in a tight loop. Updating the progress bar too frequently can slow things down.
• The code updates progress every few frames. You can tweak progress_every_n_frames in main.py.
• Some video backends report an unknown frame count (CAP_PROP_FRAME_COUNT returns 0). In that case the app can only display a status message (no accurate percentage).

Video won’t load / codec issues

• Ensure the file is a valid .mp4 (H.264 is usually the safest choice).

OpenCV window mode (cv2.imshow) doesn’t work

• In Streamlit/Docker mode the preview is rendered via st.image.
• For desktop window mode, consider switching from opencv-python-headless to opencv-python.