TL;DR

Traditional 3D reconstruction relied on iterative visual-geometry optimization (e.g., Bundle Adjustment). Recent work explored integrating machine learning via differentiable Bundle Adjustment, but remained slow and limited. VGGT (Visual Geometry Grounded Transformer) is a large feed-forward transformer that predicts all key 3D scene attributes—camera parameters, depth maps, point maps, and 3D point tracks—directly from one or many images in a single forward pass. It removes the need for geometry processing, achieves state-of-the-art results in multiple benchmarks, and runs in under a second. teaser

Background

3D reconstruction has traditionally been approached using visual-geometry methods with iterative optimization techniques like Bundle Adjustment (BA). While machine learning has played a complementary role in tasks like feature matching and depth prediction, geometry post-processing remained a crucial component, increasing complexity and computational cost. Recent approaches showed promise in direct neural reconstruction but were limited to processing only two images at once and still required post-processing for multi-view reconstruction.

Method

VGGT Achieving state-of-the-art results in multiple 3D tasks: Camera parameter estimation, Multi-view depth estimation, Dense point cloud reconstruction, 3D point tracking. Its main novelty is end-to-end training, removing the need for geometry optimization in post-processing.

Architecture

Input Processing.

  1. Takes one to hundreds of input images
  2. Images are broken down into patches that converted into tokens using DINO
  3. Appends camera tokens for camera prediction

Transformer Design.

  1. Including frame-wise and global self-attention layers
  2. Alternates between frame-wise and global self-attention layers
  3. Uses a standard transformer architecture with minimal 3D-specific inductive biases
  4. Processes all images simultaneously in a single forward pass

Output Heads.

  1. Camera head for predicting extrinsics and intrinsics
  2. DPT-based head for dense outputs (depth maps, point maps). DPT assembles multi-stage transformer tokens into image-like features at different resolutions, then uses a convolutional decoder to progressively build full-resolution predictions, benefiting from a high-resolution, globally-aware transformer backbone.
  3. Unified prediction of all 3D attributes in one pass

teaser

Training

The model is trained end-to-end on a large and diverse set of real and synthetic datasets with 3D annotations. The loss function combines supervision on multiple 3D tasks:

$$ L = L_{camera} + L_{depth} + L_{pmap} + \lambda L_{track} $$
  • Camera loss: Huber loss on camera intrinsics and extrinsics.
  • Depth loss: Aleatoric-uncertainty loss weighted by predicted confidence maps, includes pixel-wise and gradient-based terms.
  • Point map loss: Same as depth loss, but applied to predicted 3D point maps.
  • Tracking loss: Uses CoTracker-style correlation with self-attention to predict 2D correspondences across views; includes visibility estimation.

Limitations

  1. Seems difficult to train: From the paper: “We use a cosine learning rate scheduler with a peak learning rate of 0.0002 and a warmup of 8K iterations”. This is not a standard setup which indicates many iteration of training to make it work
  2. The removal of geometry supervision weakens “grounding” of the model with a greater chance of hallucination.
  3. Limited support for fisheye/panoramic imagery and failure cases under extreme rotations or strong non-rigid deformations.

Resources

Paper

Project Page

Code