I got a very low MIoU, MPA1 value when I started training. Only 0.03

[xlin]

Hello, I am very grateful to share your results. I have encountered some problems during the training process.

I briefly report my work:
1) Download data
2) Modify the loading position of the data
3) Start training
Training command: $ python train_voc.py --gpu 0

Code section: In addition to modifying the connection location of the data and printing some information about the training process, I have not modified any code.

I see that other people's training initialization MIoU is close to 0.3--0.4, and my initial is only 0.03, and there is no upward trend, I don't understand why?

[linhua]

Hello, I am very grateful to share your results. I have encountered some problems during the training process.

I briefly report my work:

1. Download data
2. Modify the loading position of the data
3. Start training
   Training command: $ python train_voc.py --gpu 0

Code section: In addition to modifying the connection location of the data and printing some information about the training process, I have not modified any code.

I see that other people's training initialization MIoU is close to 0.3--0.4, and my initial is only 0.03, and there is no upward trend, I don't understand why?

I guess you need increase you GPU nums and batch size.

[guvcolie]

@kkkmax hello! I get the same question like you, 1st epoch's mIoU is only 0.03. Do you solve it?