In this video, I am sharing out "semantic segmentation" concept for AI Model, basically some kind of ability for you to capture "layout" of your ideas in segmented color and sending it to AI, but with a bit of control of what's happening in the picture in a latent space. I don't know all the technical details, I just found it super interesting and perhaps we could run this in Blender at home and use open source tools to aid our creation.
colorseg_002.blend.zip
https://youtu.be/HPt5YiFWtLM
ControlNet #AIArt #b3d #blender3d
In this video, I am sharing out "semantic segmentation" concept for AI Model, basically some kind of ability for you to capture "layout" of your ideas in segmented color and sending it to AI, but with a bit of control of what's happening in the picture in a latent space. I don't know all the technical details, I just found it super interesting and perhaps we could run this in Blender at home and use open source tools to aid our creation.
Here is a paper: https://github.com/CSAILVision/semantic-segmentation-pytorch
Someone posted this idea on Reddit in relation to semantic segmentation and application: https://www.reddit.com/r/StableDiffusion/comments/11e4vxl/paint_by_color_numbers_with_controlnet/ https://www.reddit.com/r/StableDiffusion/comments/11flemo/color150_segmentation_colors_for_cinema4d_and/
Here is another one from the same artist, using AI to aid in packaging design: https://www.reddit.com/r/StableDiffusion/comments/115q7bq/workflow_uv_texture_map_generation_with/