Problems about the test results and the number of categories for Cityscapes

[qxr04025]

Hi, @fyu ,

Thanks for your code.
When I downloaded the Cityscapes dataset, I found the number of classes is 30， not 19 in the README.md. So I can't reproduce the mAP of the validation set. I am so confused.

And I am also confused about the difference between IOUclass and IOU category in the evaluation for Cityscapes.

Thanks.

[Pranav]

Hi @fyu,
The number of labels in Cityscapes dataset is 34 (35 if you include the -1 label of the license plate).
Attached pic is the output of the helper scripts given by Cityscape creators. Link to the script

Due to this mismatch of the number of labels, the code in which I am running a validation phase spits out the following error when it calculates the loss.

Please look into this.

[Fisher Yu]

The model is trained with 19 labels, i.e. categories with valid trainId, following the official segmentation evaluation protocal of cityscapes.