JF-Lee

I have successfully converted MTCNN and STN to onnx, but I have dimensional problems when I convert LPRNet to onnx

3

graph(%input.1 : Float(1, 3, 24, 94), %backbone.0.weight : Float(64, 3, 3, 3), %backbone.0.bias : Float(64), %backbone.1.weight : Float(64), %backbone.1.bias : Float(64), %backbone.1.running_mean : Float(64), %backbone.1.running_var : Float(64), %backbone.4.block.0.weight : Float(32,...

利用COCO训练出现问题

1

没有使用预训练模型，我在coco提取了5类70K图片进行训练，但是loss降到3.6左右就降不下来了，请问有什么策略嘛？

evaluation of problems with the model at KITTI2015

2

Hi, I try to use _torchvision.utils.save_image()_ to save the output tensor of the test, but got these blank images, and I try to convert tensor to PIL format, the image...

请问您试过将网络转onnx嘛？

2

请问在转onnx时出现这样的维度问题应该怎么解决呢？ 有一些网络改数据格式是可以解决问题的，但是这里好像不行 graph(%input.1 : Float(1, 3, 24, 94), %backbone.0.weight : Float(64, 3, 3, 3), %backbone.0.bias : Float(64), %backbone.1.weight : Float(64), %backbone.1.bias : Float(64), %backbone.1.running_mean : Float(64), %backbone.1.running_var : Float(64), %backbone.4.block.0.weight...

作者您好，请问一下在VOC0712训练一个epoch大概需要多长时间，我用2*2080Ti去训练，但是100step就需要45s, 请问我在不影响整体训练的情况下应该要如何去调整呢？

where is the source code of self.entropy_bottleneck.compress()?

2

Thanks of your excellent work! Now I wanna to deploy FactorizedPrior on embedded device, and I found EntropyBottleneck in it, but I can not find .compress() and decompress(), so can...

作者您好，请问如果我想将原始的backbone改为更轻量化的backbone，我应该如何改动呢？(看到您的arch cfg backbone中列有一系列的backbone)