Stable hyper-pooling and query expansion for event detection

Matthijs Douze 1 Jérôme Revaud 1 Cordelia Schmid 1 Hervé Jégou 2
1 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
2 TEXMEX - Multimedia content-based indexing
IRISA - Institut de Recherche en Informatique et Systèmes Aléatoires, Inria Rennes – Bretagne Atlantique
Abstract : This paper makes two complementary contributions to event retrieval in large collections of videos. First, we propose hyper-pooling strategies that encode the frame descriptors into a representation of the video sequence in a stable manner. Our best choices compare favorably with regular pooling techniques based on k-means quantization. Second, we introduce a technique to improve the ranking. It can be interpreted either as a query expansion method or as a similarity adaptation based on the local context of the query video descriptor. Experiments on public benchmarks show that our methods are complementary and improve event retrieval results, without sacrificing efficiency.
Document type :
Conference papers
Complete list of metadatas

Cited literature [18 references]  Display  Hide  Download


https://hal.inria.fr/hal-00872751
Contributor : Hervé Jégou <>
Submitted on : Monday, October 14, 2013 - 1:22:44 PM
Last modification on : Tuesday, March 12, 2019 - 5:20:03 PM
Long-term archiving on : Wednesday, January 15, 2014 - 6:10:13 AM

Files

douze_iccv13.pdf
Files produced by the author(s)

Identifiers

Citation

Matthijs Douze, Jérôme Revaud, Cordelia Schmid, Hervé Jégou. Stable hyper-pooling and query expansion for event detection. ICCV 2013 - IEEE International Conference on Computer Vision, Dec 2013, Sydney, Australia. pp.1825-1832, ⟨10.1109/ICCV.2013.229⟩. ⟨hal-00872751⟩

Share

Metrics

Record views

1750

Files downloads

1275