search

qqwweee / keras-yolo3

A Keras implementation of YOLOv3 (Tensorflow backend)
MIT License
7.14k stars 3.44k forks source link

why always 'Found 0 boxes for img' in my model tested? #553

Open sunyhaime opened 5 years ago

ChristianBlum commented 4 years ago

I had the same problem. The reason was that my detection accuracy was very low. Try to set the score in yolo.py to 0.01.

Dose someone know why my detection score is that low. I trained about 150 epochs and then there was an early stopping. The loss was still at 13. And if I test my model I will find some correct detections with the probability of e.g. 1.7% @qqwweee

ujwalv commented 4 years ago

@ChristianBlum that's a great tip, now it detects all the object, care to explain what exactly the SCORE value does ?

ChristianBlum commented 4 years ago

As far as I understood it right, with the score value you can filter out some detections with a specific detection probability. If you say score = 0.3, all the detections which have a detection probability lower than 0.3 are not shown in the picture.