Tasks

The goal of this task page is to provide you with examples of possible tasks that could be made possible with AIST Dance Video Database (AIST Dance DB).

You are welcome to contribute to this page by providing your tasks, publications, codes, and benchmark results. Please contact aistdancedb-ml@aist.go.jp so that we can maintain a list of publications that use AIST Dance DB.

Dance-motion genre classification

The dance-motion genre classification for street dances is a task of classifying 10 genres by using their video frames only.

Subset:
A subset used in this task is available at dance_genre_estimation.zip .

Results:
We investigated the accuracy of dance-motion genre classification using three-fold cross-validation (each fold used a different dancer for the test set). We first calculated the ratio of correct estimation for every dance genre and then averaged over genres to obtain the genre-classification accuracy. The best genre-classification accuracy was 91.4%.

Citation

Shuhei Tsuchida, Satoru Fukayama, Masahiro Hamasaki and Masataka Goto. AIST Dance Video Database: Multi-genre, Multi-dancer, and Multi-camera Database for Dance Information Processing. In Proceedings of the 20th International Society for Music Information Retrieval Conference (ISMIR 2019), 2019.

http://archives.ismir.net/ismir2019/paper/000060.pdf

@inproceedings{aist-dance-db,
           author = {Shuhei Tsuchida and Satoru Fukayama and Masahiro Hamasaki and Masataka Goto}, 
           title = {AIST Dance Video Database: Multi-genre, Multi-dancer, and Multi-camera Database for Dance Information Processing}, 
           booktitle = {Proceedings of the 20th International Society for Music Information Retrieval Conference, {ISMIR} 2019},
           address = {Delft, Netherlands}, 
           year = 2019, 
           month = nov }