Closed zelenooki87 closed 3 months ago
It was problem cause my VS script not using sRGB color space. Everything is OK with converted model. (tested with chainner)
However, during pytorch inference I am getting error:
python Enhancement/test_from_dataset.py --opt Options/RetinexFormer_FiveK.yml --weights pretrained_weights/FiveK.pth --dataset FiveK
export CUDA_VISIBLE_DEVICES=0
dataset FiveK
Not using Automatic Mixed Precision
===>Testing using weights: pretrained_weights/FiveK.pth
C:\Users\Miki\Retinexformer\input
C:\Users\Miki\Retinexformer\input
0%| | 0/47 [00:00<?, ?it/s]
Traceback (most recent call last):
File "C:\Users\Miki\Retinexformer\Enhancement\test_from_dataset.py", line 251, in
EDIT: only on some images.
Hi, thanks for you interests. it seems that you give three augment while three augments are needed. besides, windows system is also not suggested. Please follow the readme to install the environment and run our code.
Thank you to the author for this amazing project. I would like to convert the FiveK model to ONNX format for use in VapourSynth to process video clips. I have successfully converted the model with the following code, but the output image I get with the ONNX model is not the same as when I run the PyTorch model in the conda environment. Is there anything I need to correct in the following code for the conversion?
Thank you very much in advance. Best regards.