Attribute-Based Classification with Label-Embedding

Zeynep Akata 1, 2, * Florent Perronnin 1 Zaid Harchaoui 2 Cordelia Schmid 2
* Corresponding author
2 LEAR - Learning and recognition in vision
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : Attributes are an intermediate representation whose purpose is to enable parameter sharing between classes, a must when training data is scarce. We propose to view attribute-based image classification as a label-embedding problem: each class is embedded in the space of attribute vectors. We introduce a function which measures the compatibility between an image and a label embedding. The parameters of this function are learned on a training set of labeled samples to ensure that, given an image, the correct class has a higher compatibility than the incorrect ones. Experimental results on two standard image classification datasets are presented, resp. on the Animals With Attributes and on Caltech-UCSD-Birds datasets.
Document type :
Conference papers
Complete list of metadatas

Cited literature [12 references]  Display  Hide  Download


https://hal.inria.fr/hal-00903502
Contributor : Thoth Team <>
Submitted on : Tuesday, November 12, 2013 - 12:22:23 PM
Last modification on : Tuesday, February 12, 2019 - 10:30:05 AM
Long-term archiving on : Thursday, February 13, 2014 - 12:11:07 PM

Files

orl2013_submission_1.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00903502, version 1

Collections

Citation

Zeynep Akata, Florent Perronnin, Zaid Harchaoui, Cordelia Schmid. Attribute-Based Classification with Label-Embedding. NIPS 2013 Workshop on Output Representation Learning, Neural Information Processing Systems (NIPS) Foundation, Dec 2013, Lake Tahoe, United States. ⟨hal-00903502⟩

Share

Metrics

Record views

1280

Files downloads

909