shjo-april / recurseed_and_edgepredictmix Goto Github PK

Thanks for your impressive work.

It seems that you provide results on both deeplabv2 and deeplabv3+ in paper v1 but only deeplabv3+ result remained in paper v2. Could you provide the latest results on deeplabv2 (which most previous works adpoted) or share the code and training details of deeplabv3+ ?

Hi @OFRIN, could you please provide pseudo labels (e.g., for VOC train-aug data) used in segmentation experiments?
We just want to test them on existing deeplabV2 models. Thanks :)

Hi, could you please share your checkpoints trained on the COCO2014 dataset. Thank you!

Hi,this is a great work,It has achieved amazing results
So I look forward to your training code very much
Maybe it will be released soon?😄

Hello Sanghyun,

I'd like to thank you for making your research and resources available to the community. I've been trying to reproduce the results from your paper using the provided pseudo labels.
In your paper, it's mentioned that you achieved an accuracy of around 74.4%+ using pseudo labels. However, when I utilized your pseudo labels in conjunction with the Deeplab repo using PASCALVOC default settings, I managed to get only about 71.2+%.
I wonder if there's a specific configuration, setting, or any tweaks that I should be aware of to replicate your results accurately? Additionally, if you have a reference code or any recommended repo to train DeepLab that might be more suited to this task, it would be greatly helpful.
Looking forward to your guidance and suggestions.
Thank you for your time and assistance!

Hi, thanks for your nice work!
Would you please release the pseudo labels of COCO dataset? It will help me a lot!

Hi,
Great work so far,
we are currently building on this work. It would be great if you can release the whole code rather than a template.
Looking forward to the code.

Hi, thanks for the excellent work and looking forward to the whole code!

Where is PAMR used in the network?