Xiaodong Cun

I think web browser **DO NOT** have the ability to training the large datasets and model in one web session by the limiting of memory. convjs is great but often...

This question occurs because your image is quite large. You can try to cut the images to patches to run our method or you can just run our model in...

you need to create the dir names `logs` and `model` firstly

Hi, thanks for your attention. We report the results on ISTD because we want to compare other methods fairly. Actually, we are running some experiments on a similar dataset like...

I am sorry to reply late. There are some really busy days and I think the additional experiments will be finished in December. I didn't observe this phenomenon and can...

Hi, thanks for your attention. I just transform the images in the log space and threshold the difference. The original scripts for mask extraction are missing, however, I will try...

Hi, you can try to remove some feature layers(currently 1000+) in the VGG19 backbone, which might not influence if you reduce some of them.

Hi, thanks for your attention to our work. 1. you can try to revise the code in vgg16 hypercolumn features here,: https://github.com/vinthony/ghost-free-shadow-removal/blob/426a91026bd53cc23e8f64ed87de4fe3b4dda527/networks.py#L144-L146 Since you are working on a larger size,...

Sorry to reply late. I will update the colab soon and it is easy to modify the code by several lines.

Currently, our method might work on multiple GPUs. I am sorry i can not test your question because I do not have multiple GPU envs currently. Basically, we have implemented...