Confused about the features' name in BAIR dataset

alexlee-gk / video_prediction

Stochastic Adversarial Video Prediction

https://alexlee-gk.github.io/video_prediction/

MIT License

303 stars 65 forks source link

Confused about the features' name in BAIR dataset #20

Closed Jinghui-Wang closed 5 years ago

Jinghui-Wang commented 5 years ago

Hi! I am confused about the meaning of the features' names in BAIR dataset stored as TFRecords: 'image_aux1', 'image_main', 'endeffector_pos', 'action'. Could you give me a hint?

alexlee-gk commented 5 years ago

'image_aux1' and 'image_main' are 64x64 RGB images from two different viewpoints. 'endeffector_pos' is a 4-dimensional state vector consisting of Cartesian 3-D positions of the gripper and the rotation angle of the gripper (around the gravity axis). 'action' is a 5-dimensional vector consisting of discrete-time velocities of the state and an additional value indicating if the gripper opens or closes.

Jinghui-Wang commented 5 years ago

Got it! Thanks!