CODEWITHSUNDEEP
object-detectionyolobounding-boxdarknet
SSSSSSSSS
SSSSSSSSS

Reputation: 1

YOLOv4 multiple bounding box for the same object

I have implemented YOLOv4 object detection using darknet. After the training, I started testing the model on videos using this command : !./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -dont_show test.mp4 -i 0 -out_filename results.avi Sometimes, there are multiple bounding boxes drawn on the same object. For example, in this image example_image I have two bounding boxes. How can I display only one bounding box with the highest probability. I tried implementing IOU but failed especially that darknet contains multiple files, I didn't know which one to modify. I tried modifying darknet_video.py but with no success. PS : I tried changing the threshold but this problem persisits. Does anyone have a solution for this problem.

Upvotes: 0

Views: 2815

Answers (1)

TheGamerCoder
TheGamerCoder

Reputation: 101

Yes, it is normal. More accurately, YOLOv4 should give multiple bounding boxes per image as there multiple anchors and multi scales. For this YOLO uses Greedy-NMS(or hard-nms). It drops out boxes with less confidence and multiple boxes on one object as their IOU will be high. You should increase conf_treshold and iou_treshold, e.g. 0.5 by including the thresh flag in your command.

!./darknet detector demo cfg/coco.data cfg/yolov4.cfg yolov4.weights -dont_show test.mp4 -i 0 -out_filename results.avi -thresh 0.5

Upvotes: 2

Related Questions