Overview of the application
Human action recognition is a very interesting problem in computer vision and machine learning. There are two popular approaches to this problem,that is,still image action recognition and video action recognition. In still image action recognition, we can fine-tune a pre-trained model from ImageNet and perform a classification of the actions based on the static image. You can review the previous chapters for more information. In this chapter, we will create a model that can recognize human action from videos. At the end of the chapter, we will show you how to use multiple GPUs to speed up the training process.
Datasets
There are many available datasets that we can use in the training process, as follows:
- UCF101 (http://crcv.ucf.edu/data/UCF101.php) is an action recognition dataset of realistic action videos with 101 action categories. There are 13,320 videos in total for the 101 action categories, which makes this dataset a great choice for many research papers.
- ActivityNet...
