Output a black image

[maopian2333]

Hi, i run the inference code and finally get a black image with a warning: /opt/conda/lib/python3.9/site-packages/diffusers/image_processor.py:97: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") . It seems something wrong.

Then i change the input size to (576, 768), it gets good result without any warning.

what is the problem?

[sudip550]

Hi, i run the inference code and finally get a black image with a warning: /opt/conda/lib/python3.9/site-packages/diffusers/image_processor.py:97: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") . It seems something wrong.

Then i change the input size to (576, 768), it gets good result without any warning.

what is the problem?

does this work good with any custom images?

[levihsu]

@maopian2333 What is your original input? It seems your image format is wrong.

[maopian2333]

Hi, i run the inference code and finally get a black image with a warning: /opt/conda/lib/python3.9/site-packages/diffusers/image_processor.py:97: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") . It seems something wrong. Then i change the input size to (576, 768), it gets good result without any warning. what is the problem?

does this work good with any custom images?

@sudip550 I tried both the images in VITON-HD and the images in my own dataset, the results are the same.

[sudip550]

Hi, i run the inference code and finally get a black image with a warning: /opt/conda/lib/python3.9/site-packages/diffusers/image_processor.py:97: RuntimeWarning: invalid value encountered in cast images = (images * 255).round().astype("uint8") . It seems something wrong. Then i change the input size to (576, 768), it gets good result without any warning. what is the problem?

does this work good with any custom images?

@sudip550 I tried both the images in VITON-HD and the images in my own dataset, the results are the same.

Can you please show us your results

[maopian2333]

@maopian2333 What is your original input? It seems your image format is wrong.

@levihsu The inputs are all jpg images. Results are good when image size is resized to (576,768).

`cloth_img = Image.open(cloth_path).resize((576, 768)) model_img = Image.open(model_path).resize((576, 768)) keypoints = openpose_model(model_img.resize((384, 512))) model_parse, _ = parsing_model(model_img.resize((384, 512)))

mask, mask_gray = get_mask_location(model_type, category_dict_utils[category], model_parse, keypoints) mask = mask.resize((576, 768), Image.NEAREST) mask_gray = mask_gray.resize((576, 768), Image.NEAREST)`

[maopian2333]

@sudip550 It gets good results when input size is resized to (576,768).

It gets black images when input size is resized to (768, 1024) as original.

[maopian2333]

This problem should be caused by incorrect python package versions. I install the packages following the list you provide, then the results are good without any problem. Thanks for your work!