search

JacobYuan7 / DIN-Group-Activity-Recognition-Benchmark

[ICCV 2021] A new codebase containing various methods for Group Activity Recognition. Paper title: Spatio-Temporal Dynamic Inference Network for Group Activity Recognition.
MIT License
52 stars 15 forks source link

file tracks_normalized.pkl issue #5

Closed reacher1130 closed 2 years ago

reacher1130 commented 2 years ago

I searched all papers about group activity recognition and their source code. Your team and other teams only use this file tracks_normalized.pkl and point out where it comes from. So we don't know how to normalize the track boxes. I would appreciate it if you could know this problem.

JacobYuan7 commented 2 years ago

I searched all papers about group activity recognition and their source code. Your team and other teams only use this file tracks_normalized.pkl and point out where it comes from. So we don't know how to normalize the track boxes. I would appreciate it if you could know this problem.

Do you mean that you don't know how to normalize the boxes? I think by default, it's normalized via dividing by the length or width.

reacher1130 commented 2 years ago

I searched all papers about group activity recognition and their source code. Your team and other teams only use this file tracks_normalized.pkl and point out where it comes from. So we don't know how to normalize the track boxes. I would appreciate it if you could know this problem.

Do you mean that you don't know how to normalize the boxes? I think by default, it's normalized via dividing by the length or width.

我的意思是说在volleyball数据集中有一个默认的normalized tracks pkl文件，是因为我们都这么用，但是具体而言，pkl里面的数字是代表这什么，因为在返回数据集的py中，是将oh和ow乘以pkl中的数，我想知道其中的映射关系，谢谢！！！

JacobYuan7 commented 2 years ago

@reacher1130 里面的数字x1,y1,x2,y2就是box两个角点的坐标，它的数值范围在0-1之间是因为除以了原图的长或者宽以后的标准化到0-1之间了。

daniel-richter commented 2 years ago

Q (reacher1130):

I mean there is a default normalized_tracks.pkl file in the volleyball dataset because that's what we use, but specifically, what does the number in the pkl represent, because in the py that returns the dataset, the oh and ow are multiplied by the number in the pkl, I'd like to know the mapping relationship, thanks!

A (JacobYuan7):

The numbers x1,y1,x2,y2 are the coordinates of the two corner points of the box, and its value ranges from 0-1 because it is normalized to 0-1 after dividing by the length or width of the original figure.

Translated with www.DeepL.com/Translator (free version)

JacobYuan7 commented 2 years ago

@daniel-richter Thanks for your translation.