4xFFHQDAT

--- license: cc-by-4.0 pipeline_tag: image-to-image tags: - pytorch - super-resolution --- [Link to Github Release](https://github.com/Phhofm/models/releases/tag/4xFFHQDAT) # 4xFFHQDAT Name: 4xFFHQDAT Author: Philip Hofmann Release Date: 25.08.2023 License: CC BY 4.0 Network: DAT Scale: 4 Purpose: 4x upscaling model for faces Iterations: 122000 epoch: 2 batch_size: 4 HR_size: 128 Dataset: FFHQ - full dataset till 50k, then first 10k img multiscaled (resulted in ~260k imgs, 126GB) Number of train images: 259990 OTF Training: Yes Pretrained_Model_G: DAT_x4.pth Description: 4x photo upscaler for faces with otf jpg compression, blur and resize, trained on FFHQ dataset. This has been trained on and for faces, but i guess can also be used for other photos, might be able to retain skin detail. This is not face restoration, but simply a 4x upscaler trained on faces, therefore input images need to be of good quality if good output quality is desired. Examples 4xFFHQDAT: [Imgsli1](https://imgsli.com/MjAwNjUz) [Imgsli2](https://imgsli.com/MjAwNjU0) [Imgsli3](https://imgsli.com/MjAwNjU2) [Imgsli4](https://imgsli.com/MjAwNjU3) [Imgsli5](https://imgsli.com/MjAwNjU4) [Imgsli6](https://imgsli.com/MjAwNjU5) [Imgsli7](https://imgsli.com/MjAwNzk0)      --- Since the above 4xFFHQDAT model is not able to handle the noise present in low quality input images, i made a small variant/finetune of this, the 4xFFHQLDAT model. This model might come in handy if your input image is of bad quality/not suited for above model. I basically made this model in a response to an input image posted in upscaling-results channel as a request to this upscale model (since 4xFFHQDAT would not be able to handle noise), see Imgsli1 example below for result. Name: 4xFFHQLDAT Author: Philip Hofmann Release Date: 25.08.2023 License: CC BY 4.0 Network: DAT Scale: 4 Purpose: 4x upscaling model for low quality input p...

Source context: 0 downloads · 6 likes · Pipeline image-to-image · Repo Phips/4xFFHQDAT

Source context: 0 downloads · 5 likes · Pipeline image-to-image · Repo Phips/4xFFHQDAT