I am trying to test and train with my own images. The clothes images that I have doesn’t have the volume of the body that VITON-HD dataset has, my cloth images are plane like this: https://static.bershka.net/4/photos2/2022/I/0/1/p/8069/660/200/8069660200_2_4_1.jpg?t=1663319089433 . I have seen that the results of using an image without body volume, a plane image, are bad. VITON-HR does not recognize the back of t-shirt and put the label of the back in front. So I had to remove the back as a preprocess, but the process is tedious.
My questions are:
Has anybody train VITON-HR with plane images? Has achieve good results?
There’s a way to give a volume to those plane images?
Hello!
I am trying to test and train with my own images. The clothes images that I have doesn’t have the volume of the body that VITON-HD dataset has, my cloth images are plane like this: https://static.bershka.net/4/photos2/2022/I/0/1/p/8069/660/200/8069660200_2_4_1.jpg?t=1663319089433 . I have seen that the results of using an image without body volume, a plane image, are bad. VITON-HR does not recognize the back of t-shirt and put the label of the back in front. So I had to remove the back as a preprocess, but the process is tedious.
My questions are: