SUBIC: A supervised, structured binary code for image search
This work addresses the need for efficient image search systems by providing a supervised approach that improves accuracy over existing unsupervised methods, though it is incremental as it builds on deep learning techniques for structured quantization.
The paper tackled the problem of generating highly compressed yet meaningful binary codes for large-scale visual search by introducing a supervised, structured binary coding method using deep convolutional neural networks. The result showed that their method outperformed state-of-the-art compact representations in tasks like single and cross-domain category retrieval, instance retrieval, and classification.
For large-scale visual search, highly compressed yet meaningful representations of images are essential. Structured vector quantizers based on product quantization and its variants are usually employed to achieve such compression while minimizing the loss of accuracy. Yet, unlike binary hashing schemes, these unsupervised methods have not yet benefited from the supervision, end-to-end learning and novel architectures ushered in by the deep learning revolution. We hence propose herein a novel method to make deep convolutional neural networks produce supervised, compact, structured binary codes for visual search. Our method makes use of a novel block-softmax non-linearity and of batch-based entropy losses that together induce structure in the learned encodings. We show that our method outperforms state-of-the-art compact representations based on deep hashing or structured quantization in single and cross-domain category retrieval, instance retrieval and classification. We make our code and models publicly available online.