ReDT: Resolution-Decoupled Visual Geometry Transformer

Key points

• Existing visual geometry transformers are typically trained at fixed, standard definition (SD) resolutions and exhibit degraded reconstruction quality and poor computational scaling to high input resolutions.1
• While recent efficient architectures reduce the complexity of inter-frame attention, expensive intra-frame and encoder attention remain largely unaddressed, making HD reconstruction impractical with existing architectures.1
• By decoupling the internal representation from input resolution during attention, ReDT substantially reduces the computational cost of high-resolution 3D reconstruction and improves model performance at all resolutions.1
• Extensive experiments on multiple datasets spanning SD to FHD (2K) show that ReDT outperforms state-of-the-art visual geometry transformers in both reconstruction quality and computational efficiency.1

Sources (1)

• [1]ReDT: Resolution-Decoupled Visual Geometry Transformer
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 08:17 PM

  We introduce ReDT, a feed-forward visual geometry transformer designed to scale multi-view reconstruction to high-definition (HD) inputs.

  Existing visual geometry transformers are typically trained at fixed, standard definition (SD) resolutions and exhibit degraded reconstruction quality and poor computational scaling to high input resolutions.

Extractive summary: sentences quoted from the sources.