Loading...
Loading...
Found 3 Skills
Use this skill when implementing or evaluating image segmentation in PyTorch. Triggers include "implement semantic segmentation", "which loss function for segmentation", "compute mIoU in PyTorch", "set up a UNet with segmentation_models_pytorch", "evaluate segmentation with torchmetrics", "my Dice score is wrong", "class imbalance in segmentation", "multi-class segmentation evaluation", or "benchmark segmentation models". Handles architecture selection, loss functions, metric computation, data pipelines, and evaluation protocol. Do not use for instance segmentation benchmarks like COCO — use a separate detection skill.
SegFormer for semantic segmentation. Lightweight transformer-based architecture with hierarchical feature extraction, efficient for real-time segmentation tasks. Use when training, evaluating, exporting, quantizing, or running inference for a TAO SegFormer model. Trigger phrases include "train SegFormer", "semantic segmentation", "lightweight transformer segmenter", "real-time semantic segmentation".
Enhance an entire recording and segment it by content, preserve valuable content, remove intervals verified as invalid, and output a media library with source timecodes. Use when the user says “brighten the entire video,” “smart segmentation,” “clean up course footage,” or “preprocess and segment video”.