huntersxsx/mgpn

source code of our MGPN in SIGIR 2022

mgpn's Issues

ActivityNet dataset Problem

Problems on reproducing the scores in the paper on ActivityNet.

Dear @Huntersxsx, Thanks for your interesting work.

I have achieved similar results on Charades-sta and Tacos. However, I encountered a problem with ActivityNet.

"UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
warnings.warn("Detected call of lr_scheduler.step() before optimizer.step(). ""

To address this issue, I made the following code modification:

# state['scheduler'].step()
if state['epoch'] > 0:
    state['scheduler'].step()

However, the obtained results are as follows:

R@1,[email protected] = 46.1 (47.92 in paper)
R@1,[email protected] = 29.34 (30.47 in paper)
R@5,[email protected] = 76.26 (78.15 in paper)
R@5,[email protected] = 63.11 (63.56 in paper)

I have already set torch.backends.cudnn.deterministic = False and cudnn.benchmark = True, and I have tried many times, but the best results obtained were the ones mentioned above. Even if I ignore the warning, the performance gap still exists.

I used 4 GPUs and set the training batch size to 64 on ActivityNet. Is there anything else I should change in the code?

Looking forward to your reply.

c3d and i3d features of charades-sta

Hi!Thanks for your meaningful work.
I'm wondering where did the c3d feature and i3d features of the Charades-STA dataset come from,
Could you provide the code for extracting features or the extracted features?
Many thanks.

How to select the best result?

Hi! Thanks for your work first.
I am wondering how to select the best result during training. Is the best result chosen based on the sum of all Rank@n,mIoU@m results, or one special metric(e.g. mIoU)?