ChangeT$^3$: Evaluating Vision Test-Time Training for Remote Sensing Change Detection

Key points

• Visual test-time training offers an alternative to explicit token-to-token attention by adapting a compact inner model to each image.1
• This letter describes ChangeT$^3$, a vanilla Siamese configuration combining the existing hierarchical ViT$^3$-Tiny encoder with the multiscale ChangeFormer decoder.1
• Temporal comparison is performed by the decoder through learned fusion of paired features at four resolutions, followed by coarse-to-fine aggregation and binary prediction.1
• This separation provides a direct setting for examining image-conditioned inner updates without introducing an additional temporal interaction module.1

Sources (1)

• [1]ChangeT$^3$: Evaluating Vision Test-Time Training for Remote Sensing Change Detection
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 06:03 PM

  Bitemporal remote sensing change detection requires representations that preserve spatial detail while capturing contextual information relevant to changes between acquisitions.

  Visual test-time training offers an alternative to explicit token-to-token attention by adapting a compact inner model to each image.

Extractive summary: sentences quoted from the sources.