CNN-LSTM Architecture for Action Recognition in Videos

Action recognition in videos is currently a topic of interest in the area of computer vision, due to potential applications such as: multimedia indexing, surveillance in public spaces, among others. In this paper we propose a CNN{LSTM architecture. First, a pre-trained VGG16 convolutional neuronal n...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Orozco, Carlos Ismael, Buemi, María E., Berlles, Julio Jacobo
Formato: Objeto de conferencia
Lenguaje:Inglés
Publicado: 2019
Materias:
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/89144
Aporte de:
Descripción
Sumario:Action recognition in videos is currently a topic of interest in the area of computer vision, due to potential applications such as: multimedia indexing, surveillance in public spaces, among others. In this paper we propose a CNN{LSTM architecture. First, a pre-trained VGG16 convolutional neuronal networks extracts the features of the input video. Then, a LSTM classi es the video in a particular class. To carry out the training and the test, we used the UCF-11 dataset. Evaluate the performance of our system using the evaluation metric in accuracy. We apply LOOCV with k = 25, we obtain ~ 98% and ~ 91% for training and test respectively.