Abstract
ORAND S.A. is a Chilean company focused on developing applied research in Computer Science. This report describes the participation of the ORAND team at Instance Search task (INS) and Multimedia Event Detection task (MED) in TRECVID 2014. The INS participation consisted in three submissions to automatic detection and one submission to interactive detection. All the submissions used the four samples for each topic (type D). Our submissions were based on computing approximate k-NN search between local descriptors (without using any quantization), and computing a static similarity graph between shots to propagate scores. The MED participation consisted in two submissions to pre-specified events (010Ex and 100Ex) and two submissions to ad-hoc event detection (010Ex and 100Ex). Both submissions used the MED14EvalSub (32K videos) and noPRF. The submissions considered only low-level features (gray and color SIFT) and performed approximate k-NN searches between them without computing a codebook.
Original language | English |
---|---|
State | Published - 2020 |
Externally published | Yes |
Event | 2014 TREC Video Retrieval Evaluation, TRECVID 2014 - Orlando, United States Duration: 10 Nov 2014 → 12 Nov 2014 |
Conference
Conference | 2014 TREC Video Retrieval Evaluation, TRECVID 2014 |
---|---|
Country/Territory | United States |
City | Orlando |
Period | 10/11/14 → 12/11/14 |
Bibliographical note
Publisher Copyright:© 2020 2014 TREC Video Retrieval Evaluation. All rights reserved.