Object Detection/Identification - ComfyUI Workflow Howto

[@JotaTerrasa]

Hi there! Could you please make a video on how to use Custom Made Yolovision models?

[@AIRealm-daily]

Thanks for the video! I have a question, would be grateful if you share your thoughts. Is it possible to pair this workflow with Stable diffusion and automatically detect and inpaint certain objects, without manually outlining the inpainting area?

[@djafk]

Absolutely. The segmentation nodes have a 'mask' output that could be sent to an inpaint encode. You can grow/feather the mask in between. In the Regional Loras video I do this with people, but you could replace the 'select person' nodes with any segmenter. For example, replace a TV with a fireplace. kzbin.info/www/bejne/e3TFZ5ugmcR-prM

[@techviking23]

where can i download the workflow?

[@djafk]

I didn't make a finished workflow for that, it was just spitballing with the different modules available at the time. Since the pipelines are only ~4 nodes long it wouldn't be too hard to follow along and recreate it.

[@goodie2shoes]

those curtains.. the model should have just classified this as 'fugly'

[@djafk]

Haha, those curtains are from the 80s like the speakers.

[@suzanazzz]

Awesome video thanks for creating this! I am using Florence2 nodes for captions. and I was wondering if the caption boxes identified can change colors. Is it possible with Florence or with the nodes in this tutorial? Thanks again