Weakly-supervised Discriminative Patch Learning via CNN for Fine-grained Recognition
Authors:
Preprint
Abstract
Research on fine-grained recognition has recently shifted from multistage frameworks to convolutional neural networks (CNN) that are trained end-to-end. Many previous end-to-end deep approaches typically consist of a recognition network and an auxiliary localization network trained with additional part annotations to detect semantic parts shared across classes. To avoid the cost of extra semantic part annotations, we learn class-specific discriminative patches within the CNN framework. We achieve this by designing a novel asymmetric two-stream network architecture with supervision on convolutional filters and a non-random way of layer initialization. Experimental results show that our approach is able to find high-quality discriminative patches and achieves state-of-the-art on two publicly available fine-grained recognition datasets.Keywords
All Related Versions
This publication has 43 references indexed in Scilit:
- The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition2016
- Collaborative Layer-Wise Discriminative Learning in Deep Neural Networks2016
- Revisiting the Fisher vector for fine-grained classificationPattern Recognition Letters, 2014
- Part-Based R-CNNs for Fine-Grained Category Detection2014
- Analyzing the Performance of Multilayer Neural Networks for Object Recognition2014
- Which Looks Like Which: Exploring Inter-class Relationships in Fine-Grained Visual Categorization2014
- Jointly Optimizing 3D Model Fitting and Fine-Grained Classification2014
- Discriminative Decorrelation for Clustering and Classification2012
- Dog Breed Classification Using Part Localization2012
- Visual Recognition with Humans in the Loop2010