Anytime Dense Prediction with Confidence Adaptivity
This work addresses the need for efficient and adaptive inference in dense prediction tasks like segmentation and pose estimation, offering a novel approach that reduces computation without sacrificing accuracy.
The paper tackles the problem of anytime dense prediction, where models must produce progressively refined predictions that can be halted at any time, and introduces ADP-C, a method that achieves the same final accuracy as base models while reducing total FLOPs by 44.4% on Cityscapes semantic segmentation and 59.1% on MPII human pose estimation.
Anytime inference requires a model to make a progression of predictions which might be halted at any time. Prior research on anytime visual recognition has mostly focused on image classification. We propose the first unified and end-to-end approach for anytime dense prediction. A cascade of "exits" is attached to the model to make multiple predictions. We redesign the exits to account for the depth and spatial resolution of the features for each exit. To reduce total computation, and make full use of prior predictions, we develop a novel spatially adaptive approach to avoid further computation on regions where early predictions are already sufficiently confident. Our full method, named anytime dense prediction with confidence (ADP-C), achieves the same level of final accuracy as the base model, and meanwhile significantly reduces total computation. We evaluate our method on Cityscapes semantic segmentation and MPII human pose estimation: ADP-C enables anytime inference without sacrificing accuracy while also reducing the total FLOPs of its base models by 44.4% and 59.1%. We compare with anytime inference by deep equilibrium networks and feature-based stochastic sampling, showing that ADP-C dominates both across the accuracy-computation curve. Our code is available at https://github.com/liuzhuang13/anytime .