pic4ser / mpose2021_dataset Goto Github PK

Is there anyway to select only poses generated from videos in the utkinect dataset? I would prefer to do this through the mpose api rather than building the pose dataset on my own.

I dont understand why scale_and_center is changing my dataset:

dataset.reset_data()
dataset.get_info()
dataset.reduce_keypoints()
dataset.scale_to_unit()
dataset.add_velocities()
dataset.remove_confidence()
dataset.get_info()
dataset.scale_and_center()
dataset.get_info()

----Dataset Information----
Pose Extractor: openpose
Split: 1
X_train shape: (12562, 30, 25, 3)
X_test shape: (2867, 30, 25, 3)
Min-Max feature ranges:
x: [0.0, 828.042]
y: [0.0, 558.729]
p: [0.0, 1.0]
----Dataset Information----
Pose Extractor: openpose
Split: 1
X_train shape: (12562, 30, 15, 4)
X_test shape: (2867, 30, 15, 4)
Min-Max feature ranges:
x: [0.0, 1.0]
y: [0.0, 1.0]
Vx: [-1.0, 1.0]
Vy: [-1.0, 1.0]
----Dataset Information----
Pose Extractor: openpose
Split: 1
X_train shape: (12562, 30, 13, 5)
X_test shape: (2867, 30, 13, 5)
Min-Max feature ranges:
x: [-1.0, 1.0]
y: [-0.9995300432531916, 1.0]
Vx: [-2.0, 2.0]
Vy: [-1.9995300432531917, 1.9990533747064134]
p: [-0.9995300432531916, 1.0]

The dataset goes from having 15 keypoints to 13 keypoints and an additional channel. What am I missing here?

I am trying to run the colab notebook that you have linked. But I am just getting the keypoint values. How should I get the edge list from keypoint values?

Hello,

I want to know if it is possible to extrcat npy files from the json files extracted from a custom dataset. Is there any modification I need to apply?

Thank you for creating this repo. I would like to know the steps to prepare our custom dataset and how to create labels to use MPOSE for pose estimation dataset creation.

Good job @simoneangarano and thank you.

I just wanted to ask how do you get velocity features from the openpose pose estimation? do you subtract x2-y2 to get the velocity?

And do you have an inference.py possible or pretrained model for the purpose of testing.

Thank you very much.

In documentation string we found :param shape: desired data shape, but no shape argument in __init__