ZachZhang

How to calculate the recall rate?

3

Thanks to this open source work, I successfully calculated the mAP。 I got the following results. 49.08% = plate AP mAP = 49.08% I tried to modify the existing code...

Question about SYNTHIA dataset classes num

2

Hello~As we know, SYNTHIA dataset only share 16 classes with Cityscapes. So when training the model on the Synthia dataset, do I need to change the model classification layer to...

在FCN8模型中有一点小问题：为什么模型最终输出的shape是 （-1, nClasses）。

8

首先非常感谢分享的代码！学到很多！ 在FCN8模型中有一点小问题：为什么模型最终输出的shape是 （-1, nClasses）。理想的输出shape不应该是（batch_size,input_height, input_width,nClasses）么。因为模型提供ground truth的shape就是（batch_size, input_height, input_width, nClasses），keras在计算交叉熵的时候会自动将二者格式统一是么？如果是的话，是统一成（-1, nClasses）的格式么？具体代码如下： `def FCN8(nClasses, input_height, input_width): fcn8 = FCN8_helper(nClasses, input_height, input_width) # Conv to be applied on Pool4 skip_con1 = Conv2D(nClasses, kernel_size=(1,...

About syn_deeplab.zip

1

Hello， how can i get the syn_deeplab.zip？

My Train_loss is always 2.0~3.0

1

I have set the learn rate to 5e-7 and MAX_ITERATION to 100001, but it still does not work What should i do? Thanks!