s3_preprocess_features.py returns an empty features_X.csv and features_Y.csv

[Yaffa16]

I am trying to train the model on my own data.

The output of src/s2_put_skeleton_txts_to_a_single_txt.py is not empty.
It shows the skeleton)info.txt file as :

[[1, 1, 1, "wavingg", "wavingg_07-22-11-21-37-691/00001.jpg", 0.4176829268292683, 0.41983695652173914, 0.43597560975609756, 0.6195652173913043, 0.2926829268292683, 0.5869565217391305, 0.0, 0.0, 0.0, 0.0, 0.6067073170731707, 0.6236413043478262, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.39939024390243905, 0.3831521739130435, 0.45426829268292684, 0.37907608695652173, 0.0, 0.0, 0.5030487804878049, 0.3953804347826087], [1, 1, 2, "wavingg", "wavingg_07-22-11-21-37-691/00002.jpg", 0.4146341463414634, ......

but when I run the s3_preprocess_features.py it creates an empty festures_x.csv file.
The output reads as :

Number of samples = 128
Raw feature length = 36
Number of classes = 2
Classes: ['wavingg' 'shaking']
(128, 36)

Extracting time-serials features ...

0/128,
a new video clip starts, reset the feature generator

X.shape = (0,), len(Y) = 0
Writing features and labels to disk ...
Save features to: /home/usr/Desktop/activity_Recognition/Realtime-Action-Recognition/src/../data_proc/features_X.csv
Save labels to: /home/usr/Desktop/activity_Recognition/Realtime-Action-Recognition/src/../data_proc/features_Y.csv

Can you please help me with why the preprocess is not doing anything?

[SuChao66]

hi sir
I am trying to train the model on my own data.
The output of src/s2_put_skeleton_txts_to_a_single_txt.py is empty.
Can you help me?thanks

[felixchenfy]

@Yaffa16 Just to confirm. Step2 returns some data, but step3 is empty, correct? Looks like it's because the skeleton is missing too many parts, and in step3 (the preprocessing step) all the skeletons are discarded.

[felixchenfy]

@SuChao66 If step1 returns data, and step2 is empty, then sorry I don't know