Action-Conditioned Visual Prediction for Robot Control

Predict before execution

BridgeV2W converts robot configuration, trajectories and spatial coordinates into pixel-level dynamic masks that condition visual prediction.

Clear labeling

All future frames shown in a deployment interface are labeled as Predicted, Simulation or Generated Future. They are decision support, not guaranteed outcomes.