USC at THUMOS 2014
نویسندگان
چکیده
We submitted one run for THUMOS 2014 action recognition task. The system used improved dense trajectory features and fisher vector coding. Since testing videos are temporally untrimmed, we applied a sliding window of 100 frames for both training and testing videos. We utilized the background videos by iteratively training SVM classifiers and selecting hard negative samples. Video-level scores were generated by maximum pooling.
منابع مشابه
The LEAR submission at Thumos 2014
We describe the submission of the INRIA LEAR team to the THUMOS workshop in conjunction with ECCV 2014. Our system is based on Fisher vector (FV) encoding of dense trajectory features (DTF), which we also used in our 2013 submission. This year’s submission additionally incorporates staticimage features (SIFT, Color, and CNN) and audio features (ASR and MFCC) for the classification task. For the...
متن کاملThe THUMOS challenge on action recognition for videos "in the wild"
Automatically recognizing and localizing wide ranges of human actions are crucial for video understanding. Towards this goal, the THUMOS challenge was introduced in 2013 to serve as a benchmark for action recognition. Until then, video action recognition, including THUMOS challenge, had focused primarily on the classification of pre-segmented (i.e., trimmed) videos, which is an artificial task....
متن کاملMindLAB at the THUMOS Challenge
In this notebook paper we describe the MindLAB research group participation at the THUMOS challenge held as part of the ICCV 2013 conference. Two runs were submitted using different methods (SVM, OMF) with the features provided by the challenge organizers (DTF). The performance obtained shows an improvement over the baseline method.
متن کاملEtiquetación y desambiguación automáticas en gallego: el sistema XIADA
PhD thesis in Linguistics, written by Eva María Domínguez Noya at the University of Santiago de Compostela (USC), under the supervision of Dr. Guillermo Rojo and Dr. María Sol López Martínez (USC). The viva examination was held on the 25 of November 2013. The Examining Board consisted of the following members: Dr. Rosario Álvarez Blanco (USC), Dr. María Inês Pedrosa da Silva Duarte (University ...
متن کاملExtreme Learning Machine for Large-Scale Action Recognition
In this paper, we describe the method we applied for the action recognition task on the THUMOS 2014 challenge dataset. We study human action recognition in RGB videos through low-level features by focusing on improved trajectory features that are densely extracted from the spatio-temporal volume. We represent each video with Fisher vector encoding and additional mid-level feautures. Finally, we...
متن کامل