False positives on some classes

WongKinYiu / yolov7

Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

GNU General Public License v3.0

13.33k stars 4.2k forks source link

False positives on some classes #1334

Open faizan1234567 opened 1 year ago

faizan1234567 commented 1 year ago

I trained yolov7 on 1M million images of the person, head, car, trolley, and fall. Compared with yolov5 it gives false person and head detections while yolov5 doesn't. How to improve the model performance in those classes? It seems like it gives reasonably good accuracy in other classes. I am testing it on CCTV retail videos.

Alberto1404 commented 1 year ago

How many instances per class do you have?