Multi-Scale Dense Networks for Resource Efficient Image Classification

TL;DR

This paper introduces a multi-scale dense neural network architecture with early-exit classifiers for resource-efficient image classification, enabling anytime predictions and budgeted batch processing with improved accuracy.

Contribution

It proposes a novel multi-scale dense CNN with early-exit points, optimizing resource use and accuracy for resource-constrained image classification tasks.

Findings

Outperforms existing methods in resource-limited settings

Enables early predictions with high accuracy

Efficiently balances computation across inputs

Abstract

In this paper we investigate image classification with computational resource limits at test time. Two such settings are: 1. anytime classification, where the network's prediction for a test example is progressively updated, facilitating the output of a prediction at any time; and 2. budgeted batch classification, where a fixed amount of computation is available to classify a set of examples that can be spent unevenly across "easier" and "harder" inputs. In contrast to most prior work, such as the popular Viola and Jones algorithm, our approach is based on convolutional neural networks. We train multiple classifiers with varying resource demands, which we adaptively apply during test time. To maximally re-use computation between the classifiers, we incorporate them as early-exits into a single deep convolutional neural network and inter-connect them with dense connectivity. To facilitate high quality classification early on, we use a two-dimensional multi-scale network architecture that maintains coarse and fine level features all-throughout the network. Experiments on three image-classification tasks demonstrate that our framework substantially improves the existing state-of-the-art in both settings.