HD-CNN: Hierarchical Deep Convolutional Neural Network for Image Classification

By Zhicheng Yan, Vignesh Jagadeesh, Dennis DeCoste, Wei Di, Robinson Piramuthu. 2014

65 citations recorded by the source corpus.

Existing deep convolutional neural network (CNN) architectures are trained as N-way classifiers to distinguish between N output classes. This paper builds on the intuition that not all classes are equally difficult to distinguish from the true class label. Towards this end, we introduce hierarchical branching CNNs, called Hierarchical Deep CNN (HD-CNN), wherein classes that can be easily distinguished are classified in the higher-layer coarse category CNN, while the most difficult classifications are done on lower-layer fine category CNN. We propose utilizing a multinomial logistic loss and a novel temporal sparsity penalty for HD-CNN training. Together, they ensure each branching component deals with a subset of categories confusing to each other. Furthermore, conditional evaluation of branching components is introduced to substantially speed up the HD-CNN net execution. This new network architecture adopts a coarse-to-fine classification strategy and a module design principle. The proposed model achieves superior performance over standard models. We demonstrate state-of-the-art results on the CIFAR100 benchmark and also present preliminary results on ImageNet dataset.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Eric P. Xing
2. Xiaodan Liang
3. Zhiting Hu
4. Jianfei Cai
5. Mohit Bansal
6. Gunhee Kim
7. Jiuxiang Gu
8. Jianfeng Gao

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Dacheng Tao
3. Ming-Hsuan Yang
4. Xiaogang Wang
5. Jianfei Cai
6. G. Wang
7. Yi Yang
8. Andrew Zisserman

Citations in the corpus

• Recent advances in convolutional neural networks (2015, background): “[97] introduces a hierarchical deep CNNs (HD-CNNs) by embedding deep CNNs into a category hierarchy.”
• Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not? (2015, background): “We investigate this by training the same network with different number of individuals from 4,000 to 16,000.”
• Deep learning for visual understanding (2017, unknown): “[215] presented a hierarchical deep CNN that consists of a coarse component trained over all classes as well as several ne components trained over subsets of classes.”
• SplitNet: Learning to Semantically Split Deep Networks for Parameter Reduction and Model Parallelization (2017, background): “Yan et al. (2015) proposed a convolutional network that combines predictions from separate sub-networks for coarseand fine-grained category predictions, which share common lower-layers.”
• iPrivacy: Image Privacy Protection by Identifying Sensitive Objects via Deep Multi-Task Learning (2017, background): “For a given object class Ci , all the relevant privacy settings are re-ranked according to their object-privacy relevance scores with Ci and the privacy setting with the largest object-privacy relevance score is finally selected for the given object class Ci .”
• Deep learning for visual understanding: A review (2016, background): “For instance, [158] built on the intuition that not all classes are equally dif fi cult to distinguish from a true class label, and designed an initial coarse classi fi er CNN as well as several fi ne CNNs.”
• Poseidon: An Efficient Communication Architecture for Distributed Deep Learning on GPU Clusters (2017, background): “Deep learning (DL) is a class of machine learning (ML) approaches that has achieved notable success across a wide spectrum of tasks, including speech recognition [10], visual recognition [34, 35] and language understanding [21, 20].”
• HydraNets: Specialized Dynamic Architectures for Efficient Inference (2018, result): “HDCNN uses complex multi-stage training to mitigate optimization issues and provides robustness to routing error by overlapping the classes handled by each expert.”
• Blockout: Dynamic Model Selection for Hierarchical Deep Networks (2015, methodology): “[28, 27]) have simplified this question by fixing the base architecture and using heuristic clustering methods for separating the classes into groups.”