Box overlapping & false positive issue

[doobidoob]

Hi, thanks for providing the code and script. :-)

I trained with coco dataset using your training code. I only changed "parallels : [0,1,2,3]" to "paralles : [0]" in params.py and others are default set. Unlike what you mentioned, test results in 10, 20 epoch are not good. There are a lot of box overlapping problems and false positives as shown below. Do you know why this is happening? How can I get the detection performance shown in the script with your training code? Thank you.

[lianuo]

Haha ,I just got same problem same as yours,Thanks for asking, eagerly waiting for reply.

[lianuo]

@doobidoob what your loss value is ?

[kongshuchen]

i also meet this problem and i found that the author has changed his readme.md and remove the preformance of the model training as follow:

Results

Model	mAP (min. 50 IoU)	weights file
YOLOv3 (paper)	57.9
YOLOv3 (convert from paper)	58.18	official_yolov3_weights_pytorch.pth
YOLOv3 (our train 20 epochs)	59.66	yolov3_weights_pytorch.pth
YOLOv3 (our train 60 epochs)	61.89

His implementation of yolo v3 has lots of problems and i am try to implement another one.

[lianuo]

@kongshuchen yeah ,I tried several trainings ,it can not get good result ,hope @BobLiu20 could fixed the problem ,Thank you.

[ydixon]

Isn't it suppose to train for 100 epochs? I have only got to 10 epochs now. My recall is still at 0.47

[lianuo]

@ydixon you could test the image

[CF2220160244]

i also have this error? @doobidoob @lianuo @kongshuchen

[espectre]

i also meet this problem and i found that the author has changed his readme.md and remove the preformance of the model training as follow:

Results

Model mAP (min. 50 IoU) weights file YOLOv3 (paper) 57.9
YOLOv3 (convert from paper) 58.18 official_yolov3_weights_pytorch.pth YOLOv3 (our train 20 epochs) 59.66 yolov3_weights_pytorch.pth YOLOv3 (our train 60 epochs) 61.89
His implementation of yolo v3 has lots of problems and i am try to implement another one.

So did you implement another one?