Weird tryon results after running CP-Viton-plus on colab - am I doing something wrong!

[TamerElsherif]

Hi Matiur, @minar09
I tried to run the cp-vton-plus on Google colab but it gave weird result images in the folder "/cp-vton-plus/result/TOM/test/try-on" see below images. Is this the correct folder of the try on results? if not, where is the correct folder to look for try-on look?
can you please help me . I am still beginner in this.

Colab implementation

Result images:

Sample Images of /result/GMM/test/overlayed_TPS :

Sample Images of /result/GMM/test/result_dir :

Sample Images of /result/GMM/test/warp-cloth :

Thanks,

[minar09]

Hi @TamerElsherif , yes, the folder seems to be the correct one for the try-on results.

However, as you can see, before running the try-on (TOM) network, you need to run the warping (GMM) network first, so that the cloth is matched to the target human body. After running GMM, copy the warp-cloth and mask results to the data directory, and then run the TOM network.

Please follow the guideline carefully for your testing. Good luck!

[TamerElsherif]

Hi @TamerElsherif , yes, the folder seems to be the correct one for the try-on results.

However, as you can see, before running the try-on (TOM) network, you need to run the warping (GMM) network first, so that the cloth is matched to the target human body. After running GMM, copy the warp-cloth and mask results to the data directory, and then run the TOM network.

Please follow the guideline carefully for your testing. Good luck!

Thank you for your reply @minar09 . I did it already.
I got these results after running GMM, copying the warp-cloth and mask results folders to data and then I ran the TOM
Can you check it on Colab link below I have shared it
https://colab.research.google.com/drive/10nCk-WESMk-oDEkDIE2ATcP2AhLZ5Gqu?usp=sharing

[minar09]

Yes, your testing process seems to be okay. Is the result still same as before?

See here, cloth is not warped. You can check if the GMM pretrained model is correctly loaded or something.

[trilokpadhi]

Hi @minar09
At first , I tested the model with the test images you had given , the results are really really good , appreciate you for the awesome work that you have done
I am also trying on a custom data and getting similar results like @TamerElsherif
I have followed all the process as given by you to try on a custom data
extract the image-parse by LIPJPPNET ----> run dataset_neck_skin_connection.py ----> run body_binary_masking.py
and also generated the openpose keypoints with the COCO18 model .

During testing also , I first run the GMM model and the TOM model , but still I am not getting any reasonable result . Attached here images for your reference

final try-on result

Thanks
Trilok

[minar09]

Hi @trilokpadhi , please check why the cloth is not warped from the GMM testing. As I mentioned above, you can check whether the GMM model is correctly loaded and stuff. If it's loading correctly, can you please post here sample GMM warping results? Thanks.

[TamerElsherif]

Hi @minar09 ,

can you check the below result. I think the GMM is loaded because the GMM warped image is warped a little bit, is that correct?

What do you think i can do to make it work?

Colab instance of the project

One more question, Does the technique work for the pants, dresses, jump-suites, ans swim-suites as well?

Thanks again.

[thaithanhtuan]

Hi @TamerElsherif , can you debug by printing something to make sure that the GMM load the checkpoints correctly? It looks like it is not loaded.
Regards.

[TamerElsherif]

@thaithanhtuan , i found it wasn't able to load GMM file correctly.
thanks a lot.

[ISL270]

@thaithanhtuan , i found it wasn't able to load GMM file correctly.
thanks a lot.

what did you do to make it load the GMM file correctly?

[Muaz65]

Hi! can you please share the link of your custom image inference notebook?

[AnjaliSinghTW]

@thaithanhtuan @TamerElsherif I am having this same issue of getting colored images. Can you please help by stating the issue and its solution?