AI-Python Developer for Tool Identification Application using yolo pretrained model
Budget: ₹37,500 – ₹75,000 INR
I need a Python application that takes a pretrained YOLO model and turns it into a real-time desktop solution capable of recognising industrial, hand and power tools as they appear in a live camera feed. The workflow should be straightforward for a non-technical operator: launch the program, select the camera (USB or built-in), and immediately see bounding boxes with class names and confidence scores updating on-screen at 30 fps or better.
Key details
• Model: start with an existing YOLO checkpoint (v5, v7, v8 or YOLO-NAS—whatever you feel offers the best speed / accuracy trade-off). Feel free to fine-tune if that improves precision, but the core must stay YOLO.
• Language & libs: Python 3.x, OpenCV for video capture/rendering, torch or ultralytics for inference.
• Platform: Windows 10/11 desktop, runnable from a single installer or clean virtual environment; no cloud dependencies.
• Output: highlighted video stream plus a lightweight UI panel that shows FPS, tool class counts and a “save frame” button for manual snapshot export.
Acceptance criteria
1. Runs locally on my GPU (RTX 3060) and on CPU with acceptable fallback speed.
2. Detects the three tool categories with ≥90 % mAP on my validation clips.
3. Executable packaged (PyInstaller or similar) and full source code with README.
If you have prior experience shipping YOLO desktop apps, mention it and share a sample video or repo link—speed to demo will weigh heavily in selection.
Key details
• Model: start with an existing YOLO checkpoint (v5, v7, v8 or YOLO-NAS—whatever you feel offers the best speed / accuracy trade-off). Feel free to fine-tune if that improves precision, but the core must stay YOLO.
• Language & libs: Python 3.x, OpenCV for video capture/rendering, torch or ultralytics for inference.
• Platform: Windows 10/11 desktop, runnable from a single installer or clean virtual environment; no cloud dependencies.
• Output: highlighted video stream plus a lightweight UI panel that shows FPS, tool class counts and a “save frame” button for manual snapshot export.
Acceptance criteria
1. Runs locally on my GPU (RTX 3060) and on CPU with acceptable fallback speed.
2. Detects the three tool categories with ≥90 % mAP on my validation clips.
3. Executable packaged (PyInstaller or similar) and full source code with README.
If you have prior experience shipping YOLO desktop apps, mention it and share a sample video or repo link—speed to demo will weigh heavily in selection.