Fig 3
What the model sees
Click or → to step
Step 1 · One flat image: it fits
Camera
Depth
Joint states
More cameras
Force
1
2
3
4
Image
Depth
Joints
More cameras
Force
Image, pushed out
Sensors
Context window
1
One image
→
2
Add depth
→
3
Add joints
→
4
Add more
↺