Visual Concept Recognition and Localization via Iterative Introspection
arXiv:1603.04186
Abstract
Convolutional neural networks have been shown to develop internal representations, which correspond closely to semantically meaningful objects and parts, although trained solely on class labels. Class Activation Mapping (CAM) is a recent method that makes it possible to easily highlight the image regions contributing to a network's classification decision. We build upon these two developments to enable a network to re-examine informative image regions, which we term introspection. We propose a weakly-supervised iterative scheme, which shifts its center of attention to increasingly discriminative regions as it progresses, by alternating stages of classification and introspection. We evaluate our method and show its effectiveness over a range of several datasets, where we obtain competitive or state-of-the-art results: on Stanford-40 Actions, we set a new state-of the art of 81.74%. On FGVC-Aircraft and the Stanford Dogs dataset, we show consistent improvements over baselines, some of which include significantly more supervision.
References in corpus (12)
- Very Deep Convolutional Networks for Large-Scale Image Recognition
- Playing Atari with Deep Reinforcement Learning
- Fine-Grained Visual Classification of Aircraft
- Recurrent Models of Visual Attention
- Object Detectors Emerge in Deep Scene CNNs
- Learning Deep Features for Discriminative Localization
- Weakly Supervised Fine-Grained Image Categorization
- Neural Activation Constellations: Unsupervised Part Model Discovery with Convolutional Networks
- Fine-grained pose prediction, normalization, and recognition
- Deep Spatial Pyramid: The Devil is Once Again in the Details
- The Unreasonable Effectiveness of Noisy Data for Fine-Grained Recognition
- Self-taught Object Localization with Deep Networks