Abstract:Fine-grained classification remains a very challenging problem, because of the absence of well-labeled training data caused by the high cost of annotating a large number of fine-grained categories. In the extreme case, given a set of test categories without any well-labeled training data, the majority of existing works can be grouped into the following two research directions: 1) crawl noisy labeled web data for the test categories as training data, which is dubbed as webly supervised learning; 2) transfer the knowledge from auxiliary categories with well-labeled training data to the test categories, which corresponds to zero-shot learning setting. Nevertheless, the above two research directions still have critical issues to be addressed. For the first direction, web data have noisy labels and considerably different data distribution from test data. For the second direction, zero-shot learning is struggling to achieve compelling results compared with conventional supervised learning. The issues of the above two directions motivate us to develop a novel approach which can jointly exploit both noisy web training data from test categories and well-labeled training data from auxiliary categories. In particular, on one hand, we crawl web data for test categories as noisy training data. On the other hand, we transfer the knowledge from auxiliary categories with well-labeled training data to test categories by virtue of free semantic information (e.g., word vector) of all categories. Moreover, given the fact that web data are generally associated with additional textual information (e.g., title and tag), we extend our method by using the surrounding textual information of web data as privileged information. Extensive experiments show the effectiveness of our proposed methods.

HCL: Hierarchical Consistency Learning for Webly Supervised Fine-Grained Recognition

Hierarchical Semi-supervised Contrastive Learning for Contamination-Resistant Anomaly Detection

FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling

Fine-grained Category Discovery under Coarse-grained Supervision with Hierarchical Weighted Self-contrastive Learning

Hierarchical Fine-Grained Visual Classification Leveraging Consistent Hierarchical Knowledge

Consistency-aware Feature Learning for Hierarchical Fine-grained Visual Classification

Webly Supervised Fine-Grained Recognition: Benchmark Datasets and An Approach

Exploiting Web Images for Fine-Grained Visual Recognition by Eliminating Open-Set Noise and Utilizing Hard Examples

NCL++: Nested Collaborative Learning for long-tailed visual recognition

Hierarchical Gate Network for Fine-Grained Visual Recognition.

Label Hierarchy Constraint Network for Fine-grained Classification

Webly-Supervised Fine-Grained Visual Categorization Via Deep Domain Adaptation.

Guided by Meta-Set: A Data-Driven Method for Fine-Grained Visual Recognition

Learning Hierarchical Semantic Classification by Grounding on Consistent Image Segmentations

HmcNet: A General Approach for Hierarchical Multi-Label Classification

HCAM-CL: A Novel Method Integrating a Hierarchical Cross-Attention Mechanism with CNN-LSTM for Hierarchical Image Classification

Label Relation Graphs Enhanced Hierarchical Residual Network for Hierarchical Multi-Granularity Classification

Robust Learning from Noisy Web Images Via Data Purification for Fine-Grained Recognition.

Fine-grained Classification using Heterogeneous Web Data and Auxiliary Categories

Fine-Grained Representation Learning and Recognition by Exploiting Hierarchical Semantic Embedding