ActiveSAM Skips What Isn't There: A 5.5× Speedup for Open-Vocabulary Segmentation by Pruning Classes Before Decoding

ActiveSAM Skips What Isn't There: A 5.5× Speedup for Open-Vocabulary Segmentation by Pruning Classes Before Decoding | BedrockNews