yehao comments

Results 98 comments of


                                            yehao

Error when I tested the VGG 16 network on VOC 2007 test like the tutorial

@urumican the reason of the error is not having enough GPU memory

I found gpu_nms and cpu_nms gave different results

can you show the difference between gpu_nms and cpu_nms result?

mobilenet_v2 MAdds 429M vs paper's 300M, because of two difference

@iceflame89 good!

soft-nms cannot improve mAP used SSD detection(one-stage)

@bharatsingh430 , test results(PASCAL VOC 2007 test dataset) update: - SSD(VGG) _**nms**_(nms thresh: 0.45), mAP([email protected]): **50.2** - SSD(VGG) _**nms**_(nms thresh: 0.7), mAP([email protected]): **51.8** - SSD(VGG) _**soft-nms-linear**_(nms thresh: 0.45), mAP([email protected]): **51.3**...

soft-nms cannot improve mAP used SSD detection(one-stage)

@bharatsingh430 Thank you!

mobilenet v2-ssd?

thank you

invalid syntax: f"conv_{module_i}"

Did anyone encounter the problem?

invalid syntax: f"conv_{module_i}"

You can write like it: ![49](https://user-images.githubusercontent.com/17454232/61101611-fd1e8500-a49c-11e9-95f7-27454cde65d5.jpg) If you don't want to modify, you can use my modified version [models.txt](https://github.com/Lam1360/YOLOv3-model-pruning/files/3384919/models.txt) (tip: just modify models.txt to models.py, github don't support *.py upload)

有没有在线测试检测手的代码

你可以用我这个 [demo.txt](https://github.com/Lam1360/YOLOv3-model-pruning/files/3411408/demo.txt)

代码的实现与论文的描述一致吗？

@GUUUUG 请问你复现到了多少精度呢？