BAIR Robot Pushing

FitVid predictions on held-out videos.

• • The left image is the ground truth.

  • The right image is the predicted video.

    • The red border indicates the context frames.

    • The green border means the frame was predicted.