Choy Sum Harvesting Vision System

I need an end-to-end computer-vision pipeline that spots every Choy Sum plant in my greenhouse, decides whether it is Immature, Mature, or Overmature, pinpoints its X-Y-Z position, and flags only the harvest-ready plants for a robotic arm.

Camera & environment
• A stereo camera will be mounted above the crop beds; depth or single-lens alternatives are not planned at this stage.
• All images will come from my indoor farm, but none are yet captured or annotated, so the project starts with data acquisition and labelling.

Core objectives
• Train a YOLO-based detector in PyTorch that recognises individual plants and assigns the three growth-stage classes with high precision.
• Derive 3-D coordinates from the stereo pair, fuse them with the class prediction, and output a clean list of mature-plant positions.
• Package the results in a straightforward API or message format that my harvesting robot can consume.

What I expect to receive
- A documented data-collection protocol plus labelled dataset.
- All training scripts, inference code, and trained weights.
- A short technical report covering model architecture, evaluation metrics, and integration instructions.
- A live demo (video or remote session) that shows the system detecting, classifying, and localising plants in real greenhouse footage.

Please reply with your proposed timeline, a brief outline of your approach to stereo depth and YOLO fine-tuning, and any previous work you have done on agricultural vision systems or 3-D localisation.

---
來源：Freelancer.com | 出價次數：0 | 剩餘時間：7 日 | 專案類型：固定預算