Paper Result

[Ehteshamciitwah]

Hello,

I am training your model resnet50 on pascal (0 fold). After 70 epochs, the max MIou is 62.89. Now the MIou is Fluctuating below the maximum value. Can you suggest how I can achieve the paper results? are there any special parameters to achieve paper results?

Thank you

[juhongm999]

To reproduce the results, you have to train the model at least 350 epochs.

[Ehteshamciitwah]

Thank you. I will train until 350 epochs and let you know.

[Ehteshamciitwah]

At epoch 128, it raise an error as mentioned below. furthermore, I achieve 63.94 Miou for pascal 1shot, 0 fold on res50.

[juhongm999]

What pytorch & torchvision version do you have?

[Ehteshamciitwah]

Hello, After uploading my previous best model, I again start training and achieve paper results on resnet50 fold 0. However, I have these version Python =3.6.13 PyTorch =1.10.0 torchvision=0.11.1

Thank you for your wonderful work. I am just curious about one thing. Why the MIou of the first batch is always low during validations.

[juhongm999]

At the beginning of each epoch, mIoU is always low because the mIoU is the average over all classses and IoU results of some classes are not computed (meaning its 0 IoU) in the beginning.

[Ehteshamciitwah]

Got it. Thank you very much