The implementation may be something wrong

[espectre]

https://github.com/espectre/YOLOv3_PyTorch/blob/master/README.md
This is the result of my training.

[BCWang93]

https://github.com/espectre/YOLOv3_PyTorch/blob/master/README.md
This is the result of my training.

Hi,I have some problem use the implementation to train in the voc2007 dataset,I trained 100 epoch,but when I test the trained model use image,it has nothing!It detected is none!Do you know how to solve this?Thanks!

[espectre]

@BCWang93 His implementation may be wrong，so you could refer to the implementation of Tencent.

[BCWang93]

@BCWang93 His implementation may be wrong，so you could refer to the implementation of Tencent.

哈哈。老铁，你也在找cv的工作吗？还是已经工作的大佬？

[BCWang93]

@BCWang93 His implementation may be wrong，so you could refer to the implementation of Tencent.

大佬能给传授传授经验吗？

[espectre]

@BCWang93 Sorry, I am still a student，and maybe a rookie.

[ZhuYun97]

https://github.com/espectre/YOLOv3_PyTorch/blob/master/README.md
This is the result of my training.

Hi, I have some problem using the implementation to train in the voc2007 dataset, I trained 100 epoch, but when I test the trained model use image, it has nothing! It detected is none! Do you know how to solve this? Thanks!

I meet the same problem. I trained the model from zero on COCO dataset around 30 epochs. The loss reduced to around 0.3 and stopped falling. I have tried reduced the learning rate but got in vain. And when I test the trained model use images, it detect nothing!

[mei123hao]

https://github.com/espectre/YOLOv3_PyTorch/blob/master/README.md
This is the result of my training.

Hi, I have some problem using the implementation to train in the voc2007 dataset, I trained 100 epoch, but when I test the trained model use image, it has nothing! It detected is none! Do you know how to solve this? Thanks!

I meet the same problem. I trained the model from zero on COCO dataset around 30 epochs. The loss reduced to around 0.3 and stopped falling. I have tried reduced the learning rate but got in vain. And when I test the trained model use images, it detect nothing!

hello，now I meet the same problems, the loss reduced to 0.3 and stop falling, do you solve it ? @ZhuYun97