comfyanonymous / ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.
https://www.comfy.org/
GNU General Public License v3.0
55.51k stars 5.86k forks source link

[Feature request] Image preprocessors, e.g. HED, Canny edge, Depth #137

Open jamesWalker55 opened 1 year ago

jamesWalker55 commented 1 year ago

I would like to suggest implementing image pre-processors like HED edge detection or depth, these could process images loaded with the LoadImage node. This would be very useful for Controlnet workflows since it automates the generation of control images.

tzekeong commented 1 year ago

not sure if controlnet loader can suit your need? https://comfyanonymous.github.io/ComfyUI_examples/controlnet/controlnet_example.png

PladsElsker commented 1 year ago

someone made them as custom nodes: https://github.com/Fannovel16/comfy_controlnet_preprocessors

I think it would be awesome to have them already implemented though, as they pretty much go hand in hand with controlnets.

WASasquatch commented 1 year ago

I haven't done HED edge detection but I have depth and canny as stande-alone nodes, no installation steps, just extract and it installs what it needs: https://civitai.com/models/20793

Is HED any good? Honestly only used openpose, depth, and canny

WASasquatch commented 1 year ago

I added edge detection. It's "normal" whatever that is to PIL, and laplacian kernel.

AbyszOne commented 1 year ago

I haven't done HED edge detection but I have depth and canny as stande-alone nodes, no installation steps, just extract and it installs what it needs: https://civitai.com/models/20793

Is HED any good? Honestly only used openpose, depth, and canny

Any good? Hed is the best so far for consistency. I barely use other pp.