Shuffle rows in dataframe by specific colum value-CodePudding

I have a dataframe as follows:

  Video Frames      Feature1    Feature2  Label
0   0   frame0.jpg  feature1    feature2    0
1   0   frame1.jpg  feature1    feature2    0
2   0   frame2.jpg  feature1    feature2    0
3   0   frame3.jpg  feature1    feature2    0
4   1   frame0.jpg  feature1    feature2    1
5   1   frame1.jpg  feature1    feature2    1
6   1   frame2.jpg  feature1    feature2    1
7   1   frame3.jpg  feature1    feature2    1
8   2   frame0.jpg  feature1    feature2    0
9   2   frame1.jpg  feature1    feature2    0
10  2   frame2.jpg  feature1    feature2    0
11  2   frame3.jpg  feature1    feature2    0

I want to create slots of frames (e.g. 2 consecutive frames) and make a shuffle of rows but always ensuring that two consecutive rows of frames belonging to the same Video are kept. For the case above and slots of 2 consecutive frames, the result would be:

  Video Frames      Feature1    Feature2    Label
0   0   frame0.jpg  feature1    feature2    0
1   0   frame1.jpg  feature1    feature2    0
2   1   frame2.jpg  feature1    feature2    1
3   1   frame3.jpg  feature1    feature2    1
4   2   frame2.jpg  feature1    feature2    0
5   2   frame3.jpg  feature1    feature2    0
6   2   frame0.jpg  feature1    feature2    0
7   2   frame1.jpg  feature1    feature2    0
8   0   frame2.jpg  feature1    feature2    0
9   0   frame3.jpg  feature1    feature2    0
10  1   frame0.jpg  feature1    feature2    1
11  1   frame1.jpg  feature1    feature2    1

I want the number of frames in the slot to be configurable, i mean, for the case above i selected 2 but maybe could be 3 or 10 consecutive frames. The window of frames selected is not sliding, i mean, i select [frame0, frame1] and in the next step [frame2, frame3] but never, [frame1,frame2].

I've being thinking the best way to do it, but not clear how.

Thank you in advance

CodePudding user response：

IIUC, you can select even indices shuffle, and add the odd indices using numpy:

import numpy as np

order = np.arange(0,len(df), 2)
np.random.shuffle(order)
order = np.vstack([order, order 1]).ravel('F')

df2 = df.iloc[order]

example output:

    Video      Frames  Feature1  Feature2  Label
2       0  frame2.jpg  feature1  feature2      0
3       0  frame3.jpg  feature1  feature2      0
0       0  frame0.jpg  feature1  feature2      0
1       0  frame1.jpg  feature1  feature2      0
6       1  frame2.jpg  feature1  feature2      1
7       1  frame3.jpg  feature1  feature2      1
8       2  frame0.jpg  feature1  feature2      0
9       2  frame1.jpg  feature1  feature2      0
10      2  frame2.jpg  feature1  feature2      0
11      2  frame3.jpg  feature1  feature2      0
4       1  frame0.jpg  feature1  feature2      1
5       1  frame1.jpg  feature1  feature2      1