Apple Phenology Classification & Detection
- 1 Devlogs
- 9 Total hours
Robust deep learning models to detect images of fruit/fruitlet/flowering stages of the apple's life cycle in ochards.
trained first version of yolo world model, created two stage inference with a classification of whole image then deleting non matching predictions. strengthened bounding boxes using nms. Specialized detectors trained on one class failed. Read literature on Cascade R-CNN and other classical bounding box strengthening literature. current MAP 50:95 0.41 on validation
