Can you give me some datas to train st-gcn? about human-falling-detect-tracks HOT 15 CLOSED

I have provided create_dataset_(1-3).py files in the Data folder.

from human-falling-detect-tracks.

Can you help me how to prepare txt in create_dataset_2 @Nono17 @GajuuzZ

from human-falling-detect-tracks.

The data set is from http://le2i.cnrs.fr/Fall-detection-Dataset?lang=en but i think the download link is dead now. I wrote the script to create that data files but it quite in a manual way and specific on the data set so I didn't upload them. If you want to create your own data set you need to use the Pose model to extract the joints position (x, y, score) from each frame do some normalization and label them frame by frame then store them in a sequence of array. I wrote a comment about the inputs shape in Actionsrecognition/train.py If you have any questions you can ask.

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo. I'm doing a similar project about skeleton-based action recognition.
I was trying to train my own datasets on st-gcn, too, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.
I saw the comment in Actionsrecognition/train.py:
"Default data create from: Data.create_dataset_(1-3).py"
Are these scripts using for create the data files? If so, hoping you can provide them.
I know it is only for specific dataset, but i think if there is any example it would be easy to write my own scripts.
I will appreciate for your help.

from human-falling-detect-tracks.

Thank you!
I will keep trying on my work!

from human-falling-detect-tracks.

Thank you!
I will keep trying on my work!

can you train your own model in success？

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo.I was trying to train my own datasets on st-gcn, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.I have got Data.create_dataset_（1-3）.py.
But I don’t know what format the txt file should be in "annot_folder ='../Data/Annotation_files' # bounding box annotation for each frame." .Can you send me a copy of your data set? Let me make a reference, thank you very much!
It will be very helpful to me.
If ok ,please send me an email to [email protected].

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo.I was trying to train my own datasets on st-gcn, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.I have got Data.create_dataset_（1-3）.py.
But I don’t know what format the txt file should be in "annot_folder ='../Data/Annotation_files' # bounding box annotation for each frame." .Can you send me a copy of your data set? Let me make a reference, thank you very much!
It will be very helpful to me.
If ok ,please send me an email to [email protected].

Hi, have you known what the format of the txt file should be ? Would let me know how to make it?

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo.I was trying to train my own datasets on st-gcn, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.I have got Data.create_dataset_（1-3）.py.
But I don’t know what format the txt file should be in "annot_folder ='../Data/Annotation_files' # bounding box annotation for each frame." .Can you send me a copy of your data set? Let me make a reference, thank you very much!
It will be very helpful to me.
If ok ,please send me an email to [email protected].

Hi, have you known what the format of the txt file should be ? Would let me know how to make it?

I got it. This txt file contains a bbox of the area where the action occurs in the video.
for example:

frame label x y w h
32 0 0.5847656250000001 0.5048611111111111 0.11640625 0.47361111111111115

from human-falling-detect-tracks.

@sxy-926 Can you send me the annotation file?

@GajuuzZ, How to create own annotation file?

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo.I was trying to train my own datasets on st-gcn, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.I have got Data.create_dataset_（1-3）.py.
But I don’t know what format the txt file should be in "annot_folder ='../Data/Annotation_files' # bounding box annotation for each frame." .Can you send me a copy of your data set? Let me make a reference, thank you very much!
It will be very helpful to me.
If ok ,please send me an email to [email protected].

Hi, have you known what the format of the txt file should be ? Would let me know how to make it?

I got it. This txt file contains a bbox of the area where the action occurs in the video.
for example:

frame label x y w h
32 0 0.5847656250000001 0.5048611111111111 0.11640625 0.47361111111111115

Hi, How you created annotation files. Thank youu..

from human-falling-detect-tracks.

Hi, first i want to thanks for providing this repo.I was trying to train my own datasets on st-gcn, but i have no idea how to convert my data to proper .pkl or .npy file to fit the training scripts.I have got Data.create_dataset_（1-3）.py.
But I don’t know what format the txt file should be in "annot_folder ='../Data/Annotation_files' # bounding box annotation for each frame." .Can you send me a copy of your data set? Let me make a reference, thank you very much!
It will be very helpful to me.
If ok ,please send me an email to [email protected].

Hi, have you known what the format of the txt file should be ? Would let me know how to make it?

I got it. This txt file contains a bbox of the area where the action occurs in the video.
for example:

frame label x y w h
32 0 0.5847656250000001 0.5048611111111111 0.11640625 0.47361111111111115

Can you send me the annotation file?Thank you！！[email protected]

from human-falling-detect-tracks.

I have provided create_dataset_(1-3).py files in the Data folder.

Hello, when I run create_dataset_3.py, it's so slow with 2000 fall videos, is it normal or maybe is there some problems?
Thanks a lot! @GajuuzZ

from human-falling-detect-tracks.

I have provided create_dataset_(1-3).py files in the Data folder.

Hello, when I run create_dataset_3.py, it's so slow with 2000 fall videos, is it normal or maybe is there some problems? Thanks a lot! @GajuuzZ

May I ask how to use create_dataset_1.py to create Home_new, the labels I get using create_dataset_1.py are all 0

from human-falling-detect-tracks.

@RickChou @GajuuzZ Did you solve this problem? I am facing the same issue when I run the create_dataset_1.py. if you have solved the problem can you provide your opinion?

from human-falling-detect-tracks.