tao-mine-aoi-images

Nearest-neighbor image mining compares embeddings of source and target images to select the closest matches for augmentation. It finds source images similar to gap-analysis targets to expand real-image augmentation queues for VCN AOI training iterations.

To find nearest-neighbour source images, run the tao-mine-aoi-images skill which automates embedding target and source images, performs nearest-neighbor search, and applies label-aware filtering to select the closest matches for augmentation.

Yes, you need the tao_toolkit.data_services image dependency to perform image embedding and nearest-neighbor search for mining source images against gap-analysis targets in VCN AOI training augmentation workflows.

Yes, you can apply label-aware filtering to mined source image results. The skill includes configurable filtering to refine the nearest-neighbor matches selected for real-image augmentation queues in VCN AOI training iterations.

The best way to expand real-image augmentation queues is mining nearest-neighbour source images from a pool against gap-analysis targets. This approach uses embedding comparison to find the most similar source images for effective VCN AOI training augmentation.