BiSeNet Questions about training with customdatasets

I made the following changes： the errror:

I hope someone can help me take a look, how should I modify。Thanks！！！！！！！！！

Dec 24 '21 13:12 huangwan-jiayi

Hi,

Please make sure your input image sizes are divisible by 32.

Dec 25 '21 02:12 CoinCheung

Hi,

Please make sure your input image sizes are divisible by 32. he worked！ thank you very much！！ If I want to train a custom crop size, where should I modify the code so that the crop ratio is not restricted by 32.thank you！！

Dec 25 '21 04:12 huangwan-jiayi

Excuse me, but I encountered a new problem. When I started training and reported the following error, how can I solve it? Thank you.

Dec 25 '21 05:12 huangwan-jiayi

You can change crop size in configuration file: https://github.com/CoinCheung/BiSeNet/blob/f9231b7c971413e6ebdfcd961fbea53417b18851/configs/bisenetv1_city.py#L15

Dec 25 '21 07:12 CoinCheung

What is your dataset like? How many categories are in your dataset? What is the range of label values?

Dec 25 '21 07:12 CoinCheung

What is your dataset like? How many categories are in your dataset? What is the range of label values?

it has 3 categories ，sorry，what about the last question mean？

Dec 25 '21 08:12 huangwan-jiayi

Please check here: https://github.com/CoinCheung/BiSeNet/blob/f9231b7c971413e6ebdfcd961fbea53417b18851/lib/cityscapes_cv2.py#L67 If your label values are within (0,1,2), you need to bypass these lines about label value mapping.

Dec 27 '21 06:12 CoinCheung

Please check here:

https://github.com/CoinCheung/BiSeNet/blob/f9231b7c971413e6ebdfcd961fbea53417b18851/lib/cityscapes_cv2.py#L67

If your label values are within (0,1,2), you need to bypass these lines about label value mapping.

Do you mean to modify this file？ I made the following changes： My label map： What changes should I make to solve this problem？

Dec 27 '21 07:12 huangwan-jiayi

Have you checked every label image to make sure every label image is within [0-2] ?

Dec 28 '21 02:12 CoinCheung

There are only two categories of label images and three categories of label images in the data set

Dec 28 '21 02:12 huangwan-jiayi

I mean, have you checked one by one, and make sure each picuture's value are within [0-2]? Better try with cv2.imread(pth, 0): https://github.com/CoinCheung/BiSeNet/blob/f9231b7c971413e6ebdfcd961fbea53417b18851/lib/base_dataset.py#L53

Dec 28 '21 06:12 CoinCheung

you mean this？ The output of each label map in my data set is like the following figure。

I still haven't solved this problem。