Skip to Main content Skip to Navigation
Preprints, Working Papers, ...

PARTICUL: Part Identification with Confidence measure using Unsupervised Learning

Romain Xu-Darme 1, 2 Georges Quénot 2 Zakaria Chihani 1 Marie-Christine Rousset 3 
1 LSL - Laboratoire Sûreté des Logiciels
DILS - Département Ingénierie Logiciels et Systèmes : DRT/LIST/DILS
3 SLIDE - ScaLable Information Discovery and Exploitation [Grenoble]
LIG - Laboratoire d'Informatique de Grenoble
Abstract : In this paper, we present PARTICUL, a novel algorithm for unsupervised learning of part detectors from datasets used in fine-grained recognition. It exploits the macro-similarities of all images in the training set in order to mine for recurring patterns in the feature space of a pre-trained convolutional neural network. We propose new objective functions enforcing the locality and unicity of the detected parts. Additionally, we embed our detectors with a confidence measure based on correlation scores, allowing the system to estimate the visibility of each part. We apply our method on two public fine-grained datasets (Caltech-UCSD Bird 200 and Stanford Cars) and show that our detectors can consistently highlight parts of the object while providing a good measure of the confidence in their prediction. We also demonstrate that these detectors can be directly used to build part-based fine-grained classifiers that provide a good compromise between the transparency of prototype-based approaches and the performance of non-interpretable methods.
Complete list of metadata

https://hal-cea.archives-ouvertes.fr/cea-03703962
Contributor : Romain Xu Connect in order to contact the contributor
Submitted on : Friday, June 24, 2022 - 2:29:37 PM
Last modification on : Tuesday, August 2, 2022 - 3:10:32 AM

Files

paper.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : cea-03703962, version 1
  • ARXIV : 2206.13304

Citation

Romain Xu-Darme, Georges Quénot, Zakaria Chihani, Marie-Christine Rousset. PARTICUL: Part Identification with Confidence measure using Unsupervised Learning. 2022. ⟨cea-03703962⟩

Share

Metrics

Record views

31

Files downloads

10