HD-MTL: Hierarchical Deep Multi-Task Learning for Large-Scale Visual Recognition

In this paper, a hierarchical deep multi-task learning (HD-MTL) algorithm is developed to support large-scale visual recognition (e.g., recognizing thousands or even tens of thousands of atomic object classes automatically). To achieve more effective accomplishment of the coarse-to-fine tasks for hierarchical visual recognition, multiple sets of deep features are first extracted from the different layers of deep convolutional neural networks (deep CNNs). A visual tree is then learned by assigning the visually-similar atomic object classes with similar learning complexities into the same group, and it can provide a good environment for identifying the inter-related learning tasks automatically. By leveraging the inter-task relatedness (interclass similarities) to learn more discriminative group-specific deep representations, our deep multi-task learning algorithm can achieve the global optimum easily and obtain more discriminative node classifiers for distinguishing the visually-similar atomic object classes (in the same group) effectively. Our HD-MTL algorithm can control the inter-level error propagation effectively by using an end-to-end approach for jointly learning more representative deep CNNs (for image representation) and more discriminative tree classifier (for large-scale visual recognition) and updating them simultaneously. Our incremental deep learning algorithms can effectively adapt both the deep CNNs and the tree classifier to the new training images and the new object classes. Our experimental results have demonstrated that our HD-MTL algorithm can achieve very competitive results on both the accuracy rates and the computational efficiency for large-scale visual recognition.