-
Let's Dance: Learning From Online Dance VideosThe Let's Dance project is an effort to create a more challenging motion-heavy dataset for action recognition. We present 10 different types of dancing which are heavily dependent on motion in order to better test the parametrization of motion in videos.
-
Discovering Picturesque Highlights from Egocentric Vacation VideosWe present an approach for identifying picturesque highlights from large amounts of egocentric video data. Given a set of egocentric videos captured over the course of a vacation, our method analyzes the videos and looks for images that have good picturesque and artistic properties. We introduce novel techniques to automatically determine aesthetic features such as composition, symmetry and color vibrancy in egocentric videos and rank the video frames based on their photographic qualities to generate highlights.
-
Predicting Daily Activities from Egocentric Images Using Deep LearningIn this project we are exploring the methods in which we can analyze egocentric images to recognize activities. Thus far we have conducted a few experiments, the main one of which analyzes 40,103 labeled images (by activity) taken over a period of 6 months and uses Convolutional Neural Networks (CNNs) to classify them. We achieve an overall accuracy of 83.07% for 19 unique activities.
-
Video Segmentation ProjectThe Video Segmentation Project was an effort to bring the video segmentation and annotation algorithm to the web. We leverage the tools under Google Cloud to provide users the ability to segment and annotate their videos online. To date, over 2000 videos have been processed in the online framework.
