ML-GCN Corrected the code

Code wasn't running so i debugged the code and now its working (tested on torch version 1.7.1 )

async=True is deprecated therefor used non_blocking=True

Dec 31 '21 11:12 take2rohit

@take2rohit , hi! Have you managed to reproduce the results of the paper (or at least close to it)? I tried several times in this repository and tried to integrate this approach (ML GCN model) into my project, but the model hasn't been trained (VOC dataset, the metric is about 50-60 mAP)

Jan 11 '22 11:01 kprokofi

Hi @kprokofi . I haven't tried reproducing results. I was just using this code as a boilerplate for other code.

Jan 18 '22 07:01 take2rohit

Hi @kprokofi , did you manage to reproduce the results afterwards? Every time I train, the mAP always rise up to only 10-20.

Feb 07 '22 02:02 Akimoto-Cris

I tried to reproduce the result, but the mAP is noly about 10. Could you provide me with the training command?

Feb 12 '22 03:02 tengxiao14

https://github.com/kprokofi/ML-GCN - I couldn't reproduce the author's result, but it is got better. 93+ mAP

Feb 21 '22 09:02 kprokofi

Hi @kprokofi,

Any advise apart from the standard configs that you adopt? Appreciate advices, thanks.

Feb 21 '22 09:02 Akimoto-Cris

Gradient clipping was the bottleneck. Also, you could play with it and the learning rate.

Feb 21 '22 09:02 kprokofi

Could you provide the command for training? Thanks.

Feb 22 '22 12:02 tengxiao14

pretrained=True、lr=0.01，I got 93.4 mAP

Mar 03 '22 02:03 cym-heu

Hi, I tried to reproduce the VOC2007 project but found that if I train the model from scratch, the mAP will rise to 17 and keep still after 50 epochs. However, if I use the pre-trained model, and using the command provided from Github. The mAP will reach 90 in about 5 epochs quickly. So I wonder, how can I obtained that pretrained model, and is it possible to reproduce the author's result by training from scratch?

Mar 16 '22 07:03 LingCoder1997

excuse me ，i have tried to run this code recently but there are some problem when i run this on my computer can you help me？

Mar 19 '22 07:03 changlulu123

hi，i have tried to run this code recently but there are some problem when i run this on my computer can you help me？

Mar 19 '22 08:03 changlulu123

I was running this on the server, what kind of problem did you meet?

Mar 21 '22 01:03 LingCoderSonoscape

pretrained=True

Apr 02 '22 08:04 sorrowyn

Appreciate! Now the training process looks good!

Apr 02 '22 09:04 LingCoderSonoscape

In VOC2007, ResNet101+GMP also achieves desirable results(93.*) In MS-COCO2014, the mAP is 83.0.

Apr 02 '22 14:04 sorrowyn

Hi,

How could I load my own dataset to test with the pre-trained model?

Thanks.

Sep 30 '22 13:09 Byronliang8

Refer to this repository for more information. https://github.com/yu-gi-oh-leilei/ML-GCN_cvpr2019/blob/main/data/init.py

Sep 30 '22 13:09 sorrowyn

thanks for your help.

Oct 02 '22 12:10 Byronliang8

Hello, I am very interested in this project and would like to know how to train and test on my own data set and finally output visual results

Apr 15 '23 13:04 mjw123bs

How to set pretrained = True

Apr 22 '23 05:04 812130247

def gcn_resnet101(num_classes, t, pretrained=False, adj_file=None, in_channel=300):
    model = models.resnet101(pretrained=True) #  set pretrained = True
    return GCNResnet(model, num_classes, t=t, adj_file=adj_file, in_channel=in_channel)

Apr 23 '23 03:04 sorrowyn

Thank you so much！

Apr 23 '23 09:04 812130247

Apr 23 '23 10:04 sorrowyn

ML-GCN ML-GCN copied to clipboard

Corrected the code

ML-GCN
ML-GCN copied to clipboard