how optimize about people/characters

[zw20045500]

Thank you very much for your research work.
I’ve tested it and found that the results for people/characters are not ideal. Which inference parameter could I adjust to optimize this?
I’m looking forward to your reply. Thank you.

the script as follow:
python TVT/inferences/inference.py
--input_image /home/migu/zw/data/image/00000099.png
--output_dir /home/migu/zw/data/image-result
--pretrained_path ckp/model_TVT.pkl
--pretrained_model_name_or_path stabilityai/stable-diffusion-2-1-base
--pretrained_unet_path ckp/TVTUNet
--vae4d_path ckp/vae.ckpt
--ram_ft_path ckp/DAPE.pth
--negprompt 'dotted, noise, blur, lowres, smooth'
--prompt 'clean, high-resolution, 8k'
--upscale 2
--time_step 1 \

input image:

output image(result):

[Joyies]

Thank you very much for your research work. I’ve tested it and found that the results for people/characters are not ideal. Which inference parameter could I adjust to optimize this? I’m looking forward to your reply. Thank you.

the script as follow: python TVT/inferences/inference.py --input_image /home/migu/zw/data/image/00000099.png --output_dir /home/migu/zw/data/image-result --pretrained_path ckp/model_TVT.pkl --pretrained_model_name_or_path stabilityai/stable-diffusion-2-1-base --pretrained_unet_path ckp/TVTUNet --vae4d_path ckp/vae.ckpt --ram_ft_path ckp/DAPE.pth --negprompt 'dotted, noise, blur, lowres, smooth' --prompt 'clean, high-resolution, 8k' --upscale 2 --time_step 1 \

input image:

output image(result):

Thank you for your interest in our work. We noticed that your super-resolution setting uses a ×2 scaling factor, which is different from the ×4 real-scene super-resolution used in our paper. Since ×2 real-scene degradation is milder, it may lead to over-generation, as can be observed in your results. You may consider adjusting the degradation parameters in TVT/datasets/params_TVT.yml and retraining the model accordingly.

[Joyies]

Thank you very much for your research work. I’ve tested it and found that the results for people/characters are not ideal. Which inference parameter could I adjust to optimize this? I’m looking forward to your reply. Thank you.

the script as follow: python TVT/inferences/inference.py --input_image /home/migu/zw/data/image/00000099.png --output_dir /home/migu/zw/data/image-result --pretrained_path ckp/model_TVT.pkl --pretrained_model_name_or_path stabilityai/stable-diffusion-2-1-base --pretrained_unet_path ckp/TVTUNet --vae4d_path ckp/vae.ckpt --ram_ft_path ckp/DAPE.pth --negprompt 'dotted, noise, blur, lowres, smooth' --prompt 'clean, high-resolution, 8k' --upscale 2 --time_step 1 \

input image:

output image(result):

Thank you very much for your research work. I’ve tested it and found that the results for people/characters are not ideal. Which inference parameter could I adjust to optimize this? I’m looking forward to your reply. Thank you.
the script as follow: python TVT/inferences/inference.py --input_image /home/migu/zw/data/image/00000099.png --output_dir /home/migu/zw/data/image-result --pretrained_path ckp/model_TVT.pkl --pretrained_model_name_or_path stabilityai/stable-diffusion-2-1-base --pretrained_unet_path ckp/TVTUNet --vae4d_path ckp/vae.ckpt --ram_ft_path ckp/DAPE.pth --negprompt 'dotted, noise, blur, lowres, smooth' --prompt 'clean, high-resolution, 8k' --upscale 2 --time_step 1
input image:
output image(result):

Thank you for your interest in our work. We noticed that your super-resolution setting uses a ×2 scaling factor, which is different from the ×4 real-scene super-resolution used in our paper. Since ×2 real-scene degradation is milder, it may lead to over-generation, as can be observed in your results. You may consider adjusting the degradation parameters in TVT/datasets/params_TVT.yml and retraining the model accordingly.

In addition, we noticed that the input images are quite large. Did you apply the tile operation during inference?