This project :
1.Performs real-time object detection with YOLOv3
2.Uses OpenCV to manipulate video data
3.Uses pre-trained models and COCO dataset(of labelled images) to perform real-time and passive inference with/without a GPU
4.Develops a simple command line application with Python for inference(identification with confidence score) which saves output as a .mp4 file(yolo_utils.py&yolo.py)
Download pretrained yolov3 weights- $wget https://pjreddie.com/media/files/yolov3.weights
Download pretrained yolov3-tiny weights- $wget https://pjreddie.com/media/files/yolov3-tiny.weights
Use darknet directory: $git clone https://github.com/pjreddie/darknet
for more info on darkset visit:https://pjreddie.com/darknet/yolo/
Note:Output sample image for yolov3 and yolov3-tiny are uploaded
To run use following commands:
- for seeing yolov3-tiny output- $./darknet detector demo cfg/coco.data cfg/yolov3-tiny.cfg yolov3-tiny.weights Driving-Chinatown-SF.mp4
- for seeing yolov3 output-$./darknet detector demo cfg/coco.data cfg/yolov3.cfg yolov3.weights Driving-Chinatown-SF.mp4
- for running saving output.mp4/.mov- $python3 yolo.py -w=yolov3-tiny.weights -cfg=cfg/yolov3-tiny.cfg -v=Market\ street.mp4 -l=data/coco.names