No preview available
•Featured Project
GuavaCam: AI Guava Counting, Sizing and Weighing
YOLOOpenCVPyTorchXGBoostPython
About this Project
Computer vision system for a guava packing house. It counts, sizes and estimates the weight of fruit from camera images on a 3-lane conveyor, built during my AI Engineer role at Great Giant Foods.
Details
Three core modules on a live packing-house conveyor:
• Counting: 100% counting accuracy on the 3-lane line
• Sizing: 92% sizing accuracy, backed by camera calibration with about 1 mm error in the central region
• Weight estimation: XGBoost on 10 visual features, 87% accuracy with 17 g MAE
Detection uses YOLO segmentation (98.3% mAP@50, 97.5% recall, 96.4% F1). Figures are from my CV benchmarks.
Technologies Used
Y
YOLOO
OpenCVP
PyTorchX
XGBoostP
Python