Development of a Custom YOLO Model for Mussel Detection and Tracking

Job ID: 40343980

Budget: $10 – $30 USD

I am looking for a developer to train a custom YOLO model (YOLOv8, YOLOv11, or the newer YOLOv12/v26) specialized in detecting and tracking objects in real-time video. The primary focus is the mussel, and the model must distinguish between two specific classes: "mussel" (individual lost mussels) and "group" (clusters).

Project Requirements:

* Real-time Performance: The model will be used with a live camera feed. It must maintain at least 15 FPS on a standard NVIDIA GPU, prioritizing accuracy without sacrificing the fluid processing required for live monitoring.

* Counting & Tracking: The system must count every lost mussel per frame and maintain consistent IDs (Object Tracking) to follow individual movements over time.

* High Confidence: It must identify both classes (mussel, group) with high precision and recall.

What I will provide:

* A comprehensive dataset of over 5,000 images for labeling and training.

Deliverables:

1. Final Weights & Inference Script: Compatible with PyTorch or the Ultralytics CLI.

2. Training Documentation: A notebook or script detailing all hyperparameters used.

3. Performance Report: Metrics including Precision, Recall, mAP, and a sample video showing the tracking output.

4. README: Clear instructions to reproduce the inference and results on my end.

Once the model demonstrates accurate counts and stable ID tracking across consecutive frames, the project will be considered complete.