Fast-BoW: Scaling Bag-of-Visual-Words Generation

The bag-of-visual-words (BoW) generation is a widely used unsupervised feature extraction method for the variety of computer vision applications. However, space and computational complexity of bag-of-visual-words generation increase with an increase in the size of the dataset because of computational complexities involved in underlying algorithms. In this paper, we present Fast-BoW, a scalable method for BoW generation for both hard and soft vector-quantization with time complexities O(|h| log2 k) and O(|h|k), respectively1. We replace the process of finding the closest cluster center with a softmax classifier which improves the cluster boundaries over k-means and also can be used for both hard and soft BoW encoding. To make the model compact and faster, we quantize the real weights into integer weights which can be represented using few bits (2−8) only. Also, on the quantized weights, we apply the hashing to reduce the number of multiplications which makes the process further faster. We evaluated the proposed approach on several public benchmark datasets. The experimental results outperform the existing hierarchical clustering tree-based approach by ≈ 12 times.