HabitAction： A Video Dataset for Human Habitual Behavior Recognition

What is HHB Dataset?

HHB dataset is a fine-grained HAR dataset, which contains a large number of human habitual behaviors. The dataset contains nearly 2000 video clips, totalling nearly 300,000 frames, which are split into over 10,000 action segments, of which nearly 7000 belong to the 30 action categories.

Examples of the 30 categories of the Datasets

adjustglass

akimbo

cover face

cover mouth

cross arm

cross leg

eating fingers

fidgeting

fold hands

hand behind

hold hnee

insert pocket

lift single hand

open arms

play hairs

rest chin

roll up sleeve

rub hands

scratch head

shake legs

tickle

touch ear

touch face

touch eye

touch face

touch fore head

touch mouth

touch nose

up sweep

wipe tears

Structure of the dataset

The construction process and data structure of the proposed dataset involve the following steps: Firstly, we utilize the DC-pose model to obtain the bounding box (bbox) and skeleton information for each person in the collected videos. Then we assign action categories to each individual’s motions. In the final dataset, we have the positions of each individual, their corresponding skeleton information, and the associated action categories.

Code

View codes and instructions for reproducing the experiments on Github.

Download

The dataset can be acquired here:

https://pan.baidu.com/s/1xNuSQrIRYSa6ReMFz9vYEw?pwd=hb30

The links of the video clips are in sorting and will be released soon.

Declaration

The videos in the dataset are all aquaired from the internet, and this dataset is intended for scientific research purposes only and is not permitted for commercial use.

Page updated

Google Sites

Report abuse