Update README.md

README.md

@@ -6,9 +6,12 @@ tags:
 
 # Intro
 These are my efforts to train a real-world usable [Cascaded Gaze](https://github.com/Ascend-Research/CascadedGaze) image denoising network.
-
 denoise_util.py includes all definitions required to use Cascaded Gaze networks with PyTorch.
 
+You can find inference code for some of these models on my github:
+
+- [ComfyUI custom nodes](https://github.com/crimro-se/ComfyUI-CascadedGaze)
+
 # Models
 
 **small** (`cg_denoise_jpg+webp_artifacts_small.safetensors`)
@@ -44,6 +47,12 @@ model.requires_grad_(False)
 model.eval()
 ```
 
+**sidd** (`cg_sidd.safetensors`)
+The official SIDD benchmark trained CascadedGaze model. I have ported the weights and added metadata such that it can be used as easily as my small model.
+
+However, my view is that the SIDD dataset is poor, and as a result this model is not useful in any task.
+
+
 **v1**
 - an early experiment, not recommended
 - ~ 132M params, trained on 256 * 256 **RGB** patches for intermediate jpg & webp compression artefact removal. It's been trained on about 700k samples (photographs only) at a precision of bf16. Also capable of removing ISO-like noise and gaussian noise.