A Swin transformer (Swin-BIDDS) beats ResNet and ViT baselines for long-term body identification, and larger 384 by 384 pixel input images account for most of the gain.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Unconstrained Body Recognition at Altitude and Range: Comparing Four Approaches
A Swin transformer (Swin-BIDDS) beats ResNet and ViT baselines for long-term body identification, and larger 384 by 384 pixel input images account for most of the gain.