Project for detecting and classify Vulnerable Road Users (VRU) on road.

Image Processing Utilities

Crop Images

Cropping out objects and reshape them into the suitable size for the neural network

1. Modify global variables RAW_DATA_PATH, PROCESSED_TRAIN_DATA_PATH, PROCESSED_VAL_DATA_PATH, JSON_F, NEW_SHAPE and TRAIN_TO_VAL_RATIO in file process_img_from_json.py
2. Run python process_img_from_json.py

Detect Objects in Images

Detect objects with YOLO v3 object detection model

1. Put images to be processed in <IMAGE_DIR>
2. Run python obj_det/det.py --images <IMAGE_DIR> --det <OUTPUT_DIR>

Merge Bounding Boxes

Find bounding box groups that are potentially wheelchair users and merge the bounding boxes

1. Run python test.py