I have a some question about pascal voc train

[dypark86]

Thanks for your jobs!

I have a some question about train method I train pascal voc dataset model with basic setting. and evaluate model with eval_all.py.

but there are some gap in the performance.

• my model 87.2
• your model 90.5
• 3.3 gap

there are some more training technic?

• i train just 100 epoch.
• more train epoch needs?

thanks!

[kmaninis]

Hi,

Could you provide some more info about the setup you are using? Python/PyTorch version, GPU, etc.? The number should be reproducible just by running the script, without changing any parameters (tripple checked).

[dypark86]

i tried with following setup

• pytorch version 0.4.0
• 1 gpu, Titan V
• 5 batches

after setting the data path i edit just following code -> before: device = torch.device("cuda:"+str(gpu_id) if torch.cuda.is_available() else "cpu") -> after: device = torch.device("cuda") because setgpuid make error in my env

and i train with train_pascal.py and eval with eval_all.py

Thanks!

[scaelles]

Hello, We tried with the same configuration except that we are using a TItan Xp, but it shouldn't make a difference. Could you try to modify the threshold in eval_all.py? Best, Sergi