Clustering web video search results with convolutional neural networks

Convolutional Neural Networks (CNNs) have been established as a powerful class of models for image recognition problems giving state-of-the-art results on recognition, detection, segmentation, classification and retrieval. Encouraged by these results, we develop our previous work [14] by implementing deep neural network architecture for extracting and representing visual features to improve the clustering quality of web video search results. Experiments were conducted on a dataset published in [14]. This dataset includes 1580 videos from 18 queries issued to the YouTube search engine. Our method exhibits significant performance improvements over the previously published result evaluated by Entropy measure (23.27% vs. 39.46%) and Purity measure (77.09% vs. 61.50%).