AdaPool: Exponential Adaptive Pooling for Information-Retaining Downsampling

TL;DR

AdaPool introduces an adaptive, exponential pooling method that enhances detail preservation in CNNs and offers a bidirectional approach for both downsampling and upsampling, improving performance across various vision tasks.

Contribution

The paper proposes adaPool, a novel adaptive pooling technique that learns regional-specific fusion of kernels based on exponential functions, with a bidirectional property for downsampling and upsampling.

Findings

AdaPool outperforms traditional pooling methods on image and video classification tasks.

AdaPool improves detail preservation in super-resolution and frame interpolation.

The method introduces minimal additional computational and memory overhead.

Abstract

Pooling layers are essential building blocks of convolutional neural networks (CNNs), to reduce computational overhead and increase the receptive fields of proceeding convolutional operations. Their goal is to produce downsampled volumes that closely resemble the input volume while, ideally, also being computationally and memory efficient. Meeting both these requirements remains a challenge. To this end, we propose an adaptive and exponentially weighted pooling method: adaPool. Our method learns a regional-specific fusion of two sets of pooling kernels that are based on the exponent of the Dice-Sorensen coefficient and the exponential maximum, respectively. AdaPool improves the preservation of detail on a range of tasks including image and video classification and object detection. A key property of adaPool is its bidirectional nature. In contrast to common pooling methods, the learned weights can also be used to upsample activation maps. We term this method adaUnPool. We evaluate adaUnPool on image and video super-resolution and frame interpolation. For benchmarking, we introduce Inter4K, a novel high-quality, high frame-rate video dataset. Our experiments demonstrate that adaPool systematically achieves better results across tasks and backbones, while introducing a minor additional computational and memory overhead.