TRAIN ON CUSTOM DATASET | SEGMENTATION

[EthanAbitbol3]

Hello and thank you for sharing your work, i was wondering if you woul dprovide some information about training on custom dataset ?

Thank you !

[fschvart]

You can look at mmsegmentation 0.27 repo demo. You'll see how to set up a custom dataset there

[EthanAbitbol3]

I appreciate your help! Prior to attempting the method you suggested, I also made an effort to retrain using the ADE20k dataset to evaluate its performance. Regrettably, it did not yield the desired results.

Here is the command I executed: !set CUDA_VISIBLE_DEVICES=0 !python train.py configs/ade20k/upernet_internimage_t_512_160k_ade20k.py

However, I encountered the following error:

I proceeded to make the following modifications to the upernet_internimage_t_512_160k_ade20k.py file:

Additionally, I made the following changes to the ade20k.py file:

I encountered the same problem again with the cfg.dump function, which caused the issue to persist.

NB : I am on Windows.

[fschvart]

Shalom @EthanAbitbol3 ,

Even though you're on Windows, use relative reference and '/' instead of '' or '\'

[EthanAbitbol3]

Shalom! Thank you for your response. I have made the changes as you suggested.

Unfortunately, I encountered the same error once again.