Zeynep Akata , Xerox Research Centre Europe, Meylan and INRIA Rhone-Alpes, Montbonnot
Florent Perronnin , Xerox Research Centre Europe, Meylan
Zaid Harchaoui , INRIA Rhone-Alpes, Montbonnot
Cordelia Schmid , INRIA Rhone-Alpes, Montbonnot
We benchmark several SVM objective functions for large-scale image classification. We consider one-vs-rest, multi-class, ranking, and weighted approximate ranking SVMs. A comparison of online and batch methods for optimizing the objectives shows that online methods perform as well as batch methods in terms of classification accuracy, but with a significant gain in training speed. Using stochastic gradient descent, we can scale the training to millions of images and thousands of classes. Our experimental evaluation shows that ranking-based algorithms do not outperform the one-vs-rest strategy when a large number of training examples are used. Furthermore, the gap in accuracy between the different algorithms shrinks as the dimension of the features increases. We also show that learning through cross-validation the optimal rebalancing of positive and negative examples can result in a significant improvement for the one-vs-rest strategy. Finally, early stopping can be used as an effective regularization strategy when training with online algorithms. Following these "good practices", we were able to improve the state-of-the-art on a large subset of 10K classes and 9M images of ImageNet from 16.7% Top-1 accuracy to 19.1%.
Image Processing and Computer Vision, Sparse, structured, and very large systems, Gradient methods, Stochastic programming, Clustering, classification, and association rules, Information Search and Retrieval, Computer vision, Machine learning
F. Perronnin, Z. Akata, Z. Harchaoui and C. Schmid, "Good Practice in Large-Scale Learning for Image Classification," in IEEE Transactions on Pattern Analysis & Machine Intelligence.