Computer Vision

Core Innovation: Linear-Time 3D Reconstruction

Feed-forward transformer models have driven rapid progress in 3D vision, but state-of-the-art methods such as VGGT and π³ have a computational cost that scales quadratically with the number of input images, making them inefficient when applied to large image collections. Sequential-reconstruction approaches reduce this cost but sacrifice reconstruction quality. We introduce ZipMap, a stateful feed-forward model that achieves linear-time, bidirectional 3D reconstruction while matching or surpassing the accuracy of quadratic-time methods. ZipMap employs test-time training layers to zip an entire image collection into a compact hidden scene state in a single forward pass, enabling reconstruction of over 700 frames in under 10 seconds on a single H100 GPU—more than 20× faster than SOTA methods such as VGGT. Moreover, we demonstrate the benefits of having a stateful representation in real-time scene state querying and its extension to sequential streaming reconstruction. Project: https://haian-jin.github.io/ZipMap

Key Innovations:

• 3D Reconstruction: Efficiently reconstructs 3D scenes from images or videos.
• Linear-Time: Achieves computational scaling linear with the number of input images, vastly improving efficiency.
• Test-Time Training: Employs TTT layers to compress image collections into a compact hidden scene state.
• Computer Vision: Advances the state-of-the-art in 3D perception and scene understanding.
• Deep Learning: Leverages advanced deep learning architectures for enhanced performance.
• Stateful Models: Utilizes a stateful representation for real-time querying and streaming reconstruction.