Hand Detection in Egocentric Video and Investigation Towards Fine-Grained Cooking Activities Recognition
Katsufumi Inoue, Misa Ono, Michifumi Yoshioka · Information Engineering Express · 2019
The analysis of egocentric videos is currently a hot topic in computer vision.In this paper, we focus on cooking activities recognition in egocentric videos.To recognize cooking activities automatically and precisely, we must solve the problems of detecting hand region in egocentric videos, representing hand motion, and classifying the cooking activities.In this research, to solve these problems, we propose a new cooking activities recognition method in egocentric videos.The characteristic points are 1) hand regions are accurately detected in a cluttered background by using color, texture, and location information, 2) temporal hand features are extracted from sequential frame images with a thinning algorithm, 3) a fully-connected multi-layer neural network is utilized to recognize the activities from the extracted features.Toward our goal of fine-grained cooking activities recognition, we investigated the performance of our method with our benchmark, including 12 fine-grained cooking activities in five coarse categories.The experimental results show that our method allows us to recognize cooking activities with an accuracy of 45.2%.