Train Mask2Former to segment your images.
Trains, evaluates, exports, quantizes, and runs Mask2Former models for panoptic, instance, or semantic image segmentation.
When to use it
Use for training, evaluation, export, quantization, or inference requests involving TAO Mask2Former.
Give it segmentation datasets or model files and a requested model action; it runs the corresponding workflow and saves results to an output directory.
What you provide
Requires Docker for the container-based workflow.
Requires NVIDIA Container Toolkit alongside Docker.
This skill
GitHub
Downloads Swin backbone weights
/ptm/mask2former/swin_tiny_patch4_window7_224_22k/swin_tiny_patch4_window7_224_22k.pth
Saves pretrained backbone weights