question about the data

[liangyuxin42]

Thanks for sharing this great work!
I have some questions about the data , I am reading the data through the following code:

h5 = h5py.File("/mobile_aloha/aloha_mobile_cabinet/episode_2.hdf5", 'r')
action = h5['/action'][()]
print(action.shape) # (1500, 14)

1. is this 14 DoF = 2 arms * (6 joint positions + 1 gripper)?
2. if so, is there some way to calculate end-effector positions from joint positions in order to convert the data into a similar form as RT-X data (7-dimensional action vector controlling the end-effector: x, y, z, roll, pitch, yaw, and gripper opening)
   I want to mix different robot data for training, so unifying the different formats would be very useful.

Thanks!

[asasarkaya]

@liangyuxin42 I am using the same link to get data, and the images are not three dimensional. Do you have any idea about this? I would appreciate any info. Here are some samples shapes of the images from ziploc data (on the other hand, when I record simulation data, they are recorded & loaded as 480,640,3, so they are fine) :

sample_image_cam_high: (22229,)
sample_image_cam_low: (24912,)
sample_image_cam_left_wrist: (17664,)
sample_image_cam_right_wrist: (18395,)

[liangyuxin42]

@asasarkaya this is how i read the image

import h5py
import io
from PIL import Image
import numpy as np

h5 = h5py.File("/episode_2.hdf5", 'r')
i = 10
data = np.array(h5["/observations/images/cam_right_wrist"][:][i])
img = Image.open(io.BytesIO(data))
print(img.size)
img

the output is :
(640, 480)

hope this helps!

[asasarkaya]

thanks!