Update app.py

app.py

@@ -405,93 +405,4 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange")) as iface:
 
 if __name__ == "__main__":
     iface.launch()
-```
-
-### Key Changes
-1. **Performance**:
-   - Removed `shutil.rmtree` to reduce I/O overhead; directories are recreated if needed.
-   - Added per-frame timing logs (`Frame {frame_count}: Processed in {frame_time:.2f} ms`) to diagnose bottlenecks.
-   - Introduced a 600-second timeout to prevent runaway processing, logging if exceeded.
-   - Fixed `frame_skip=5` as default, with slider range of 1–5 to prevent accidental full-frame processing.
-
-2. **Download**:
-   - Kept `gr.File` components for downloading `processed_output.mp4`, `data_lake_submission.json`, `captured_frames.zip`, `flight_logs.zip`, `chart_temp.png`, `map_temp.png`.
-   - ZIP creation runs post-processing to avoid runtime overhead.
-
-3. **Codec**:
-   - Uses only `mp4v` for MP4 output, logging errors to ensure compatibility.
-
-4. **Logging**:
-   - Added frame-level timing and input video details for debugging.
-   - Fixed minor logging errors (e.g., removed duplicate `Output video` print).
-
-5. **NHAI Compliance**:
-   - Imagery: Geotagged JPEGs in `captured_frames.zip`. Compliant at 4000x3000; non-compliant at 768x432.
-   - Flight Logs: CSVs in `flight_logs.zip`. Compliant.
-   - Data Submission: `data_lake_submission.json`. Compliant.
-   - Analytics: Limited to 3 classes. Partially compliant.
-   - Video/Charts/Maps: Optional, compliant.
-
-### Why the Slowdown?
-The most likely cause is a higher-resolution or longer input video (e.g., 12 MP or >633 frames). To confirm:
-- Check the input video’s properties in the logs (`Input video: ...`).
-- If it’s 768x432 with 633 frames, processing should take ~146 seconds.
-- If it’s 12 MP (e.g., 4000x3000), expect ~5250 seconds on CPU for 633 frames.
-- Verify `frame_skip` in the Gradio interface (should be 5).
-
-### How to Test
-1. **Setup**:
-   - Update your Space with the new `app.py` (above).
-   - Ensure `./data/best.pt` is uploaded.
-   - Use `requirements.txt`:
-     ```text
-     torch
-     ultralytics
-     gradio==5.33.2
-     numpy
-     opencv-python
-     matplotlib
-     piexif
-     ```
-   - Confirm ~10 GB storage.
-
-2. **Test with 768x432 Video**:
-   - Upload your 768x432 video (633 frames).
-   - Set sliders to 768x432 (faster) or 4000x3000 (default, slower).
-   - Ensure `frame_skip=5`.
-   - Click “Process Video” (~146 seconds expected).
-   - Verify:
-     - No errors in logs.
-     - Processing completes in ~146–200 seconds.
-     - Video, metrics, logs, gallery, chart, map display.
-     - Download buttons for:
-       - `data_lake_submission.json`
-       - `captured_frames.zip` (~35 images)
-       - `flight_logs.zip` (CSVs)
-       - `processed_output.mp4`
-     - Logs show frame timings, non-12 MP warnings.
-   - Check `app.log` for frame-level timings if issues persist.
-
-3. **Test with 12 MP Video**:
-   - Upload a 12 MP video (e.g., 4000x3000).
-   - Set sliders to 4000x3000, `frame_skip=5`.
-   - Expect longer processing (~5000 seconds on CPU).
-   - Confirm downloads and no warnings.
-
-### Notes
-- **Input Video**:
-  - Please confirm the current video’s resolution, frame count, and duration. If it’s not 768x432 with 633 frames, the slowdown is expected.
-  - Use `ffmpeg -i input_video.mp4` or check logs for details.
-
-- **Performance**:
-  - CPU is slow for 12 MP videos. Enable GPU in Space settings.
-  - If `frame_skip=1`, set it back to 5 for faster processing.
-
-- **NHAI Compliance**:
-  - 768x432 is non-compliant (12 MP required). Use 12 MP for production.
-  - Retrain model for Annexure-III’s 18 classes.
-
-- **Codec**:
-  - If codec warnings persist, install `opencv-python-headless` with FFmpeg support.
-
-If processing is still slow (>200 seconds for 768x432), share the latest logs or video details (resolution, frames), and I’ll debug further!
+