Towards Flexible Classification: Cost-Aware Online Query of Cascades and Operating Points

We propose a method that uses off-the-shelf binary classifiers and combines them in real-time to achieve quality, time, and cost characteristics that satisfy user-provided constraints; moreover, it shares intermediate computation (e.g., features, classifiers, kernels) between classes to reduce time and cost. An important distinction between this approach and others is that the primary focus of this work is on improving flexibility, operation on a wide range of operating points, as opposed to quality or speed; however, we show that the proposed approach achieves performance comparable to the state-of-the-art on the scene recognition task for the SUN397[1] where we attain a mean AUC of 0.9625 compared to 0.9573[1], 0.8426 [2], 0.9227[3], and 0.9343(based on [4]). Additionally, we show that human annotators can be naturally integrated into the approach to produce hybrid human/algorithmic classifiers.