The Mask Segmentation tab is used to generate masks for images (masks for videos are currently not supported).
- Masks are optional but can for example be used with OneTrainer and its masked training features.
- Masks are generated using the Segment Anything 3 Model (SAM3). Please see the readme_models.md for more information on how to download and install models.
- To generate masks start by defining the "Prompt", this will be the subject you want to mask. For example "cat", or "person", or "woman" etc.
- You can optionally change the other settings, they have tooltips to explain what they do.
- Hit "START PROCESSING" to process all files in your folder.
- For each image file a mask will be generated and saved in the same folder as the image file with the same name but with "-masklabel" appended to the filename and in the PNG file format.
- The "Separate File (*-masklabel.png) file naming scheme is directly compatible with OneTrainer.
- For other training tools you may want to experiment with the other file naming schemes.