Abstract
Convolutional neural networks excel in a number of computer vision tasks. One of their most crucial architectural elements is the effective receptive field size, which has to be manually set to accommodate a specific task. Standard solutions involve large kernels, down/up-sampling and dilated convolutions. These require testing a variety of dilation and down/up-sampling factors and result in non-compact networks and large number of parameters. We address this issue by proposing a new convolution filter composed of displaced aggregation units (DAU). DAUs learn spatial displacements and adapt the receptive field sizes of individual convolution filters to a given problem, thus reducing the need for hand-crafted modifications. DAUs provide a seamless substitution of convolutional filters in existing state-of-the-art architectures, which we demonstrate on AlexNet, ResNet50, ResNet101, DeepLab and SRN-DeblurNet. The benefits of this design are demonstrated on a variety of computer vision tasks and datasets, such as image classification (ILSVRC 2012), semantic segmentation (PASCAL VOC 2011, Cityscape) and blind image de-blurring (GOPRO). Results show that DAUs efficiently allocate parameters resulting in up to 4× more compact networks in terms of the number of parameters at similar or better performance.
Original language | English |
---|---|
Pages (from-to) | 2049-2067 |
Number of pages | 19 |
Journal | International Journal of Computer Vision |
Volume | 128 |
Issue number | 8-9 |
Early online date | 2 Jan 2020 |
DOIs | |
Publication status | Published - Sept 2020 |
Keywords
- Adjustable receptive fields
- Compact ConvNets
- Displacement units
- Efficient ConvNets
ASJC Scopus subject areas
- Software
- Computer Vision and Pattern Recognition
- Artificial Intelligence