search

andy-yun / pytorch-0.4-yolov3

Yet Another Implimentation of Pytroch 0.4.1 and YoloV3 on python3
MIT License
279 stars 72 forks source link

About the predicted result #39

Closed lilyzhuang closed 5 years ago

lilyzhuang commented 5 years ago

Hi, Andy. I have two questions about the output.

  1. I found that there is something wrong in the detected box. After using the function correct_yolo_boxes( ), the left, top of the box can be negative.

e.g. -3.15759015083 26.1544475555 92.0868988037 106.778274536

  1. The conf_threshold used in testing is 0.5 while it is set to 0.25 in training. I think they should be the same.

Another question is about non-maximum suppression in training. I didn't find it in the training procedure. Do you use nms in the training procedure?

Many thanks.

andy-yun commented 5 years ago
  1. The box might be detected at the out of image. But, I don't think that it is a bug. ^__^.
  2. nPP results are shown over the value 0.25. You're right. But, sometimes, if you set a high value to threshold of nPP, you can see nPP = 0 very long time. Thus, I set the low value.
andy-yun commented 5 years ago

@LZZZZ29 you may got the error when the text box at the result is placed at out of image. When the case, drawtext at utils.py shows exception error (ValueError). I fixed that and uploaded new utils.py.

lilyzhuang commented 5 years ago

Sure. There is a problem in the location of the text box. Many thanks for your update. @andy-yun

mrkieumy commented 5 years ago

Hi @andy-yun and @LZZZZ29, sorry for asking an old question. I downloaded the newest utils.py but I don't know why the center coordinates still have minus value (-10,...). I do not understand it because I think center x and y calculate based on image size and it must be positive value right? Do you have any idea to solve it? Thanks & best regards.