Unsupervised Adversarial Instance-level Image Retrieval - CNRS - Centre national de la recherche scientifique Accéder directement au contenu
Article Dans Une Revue IEEE Transactions on Multimedia Année : 2021

Unsupervised Adversarial Instance-level Image Retrieval

C. Bai
  • Fonction : Auteur correspondant
  • PersonId : 1095757

Connectez-vous pour contacter l'auteur
H. Li
  • Fonction : Auteur
  • PersonId : 1095758
J. Zhang
  • Fonction : Auteur
  • PersonId : 1095759

Résumé

With the wide use of visual sensors in the Internet of Things (IoT) in the past decade, huge amount of images are captured in people's daily lives, which makes the traditional deep learning-based image retrieval frameworks meeting some challenges. For one, most of them need a large amount of annotated training data, which is expensive. Moreover, machines still lack human intelligence, as illustrated by the fact that they pay less attention to the interesting regions that humans generally focus on when search images. Hence, this paper proposes a novel unsupervised framework that focuses on the instance object in the image, and integrates human intelligence into the deep learningbased image retrieval. This framework is called adversarial instance-level image retrieval (AILIR). We incorporate adversarial training and an attention mechanism into this framework that consider human intelligence with artificial intelligence. The generator and discriminator are redesigned to guarantee that the generator retrieves similar images while the discriminator selects unmatched images and creates an adversarial reward for the generator. A minimax game is conducted by the adversarial reward retrieval mechanism until the discriminator is unable to judge whether the image sequence retrieved matches the query. Comparison and ablation experiments on four benchmark datasets prove that the proposed adversarial training framework indeed improves instance retrieval and outperforms the state-ofthe-art methods focused on instance retrieval.
Fichier principal
Vignette du fichier
Bai et al-2021-Unsupervised Adversarial Instance-level Image Retrieval.pdf (9.3 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03195862 , version 1 (30-06-2021)

Identifiants

Citer

C. Bai, H. Li, J. Zhang, L. Huang, Lu Zhang. Unsupervised Adversarial Instance-level Image Retrieval. IEEE Transactions on Multimedia, 2021, 23, pp.2199-2207. ⟨10.1109/TMM.2021.3065578⟩. ⟨hal-03195862⟩
93 Consultations
145 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More