Researchers found that standalone vision transformer models deliver higher segmentation accuracy and lower latency than ...