Expanding-and-Shrinking Binary Neural Networks
This work addresses the problem of low accuracy in BNNs for applications requiring efficiency, offering a method that improves performance while maintaining computational benefits, though it appears incremental as it builds on existing binarization techniques.
The paper tackles the accuracy degradation of binary neural networks (BNNs) in challenging tasks by proposing an expanding-and-shrinking operation to enhance binary feature maps with minimal computational overhead, achieving remarkable improvements over leading binarization algorithms across diverse applications like image classification and object detection.
While binary neural networks (BNNs) offer significant benefits in terms of speed, memory and energy, they encounter substantial accuracy degradation in challenging tasks compared to their real-valued counterparts. Due to the binarization of weights and activations, the possible values of each entry in the feature maps generated by BNNs are strongly constrained. To tackle this limitation, we propose the expanding-and-shrinking operation, which enhances binary feature maps with negligible increase of computation complexity, thereby strengthening the representation capacity. Extensive experiments conducted on multiple benchmarks reveal that our approach generalizes well across diverse applications ranging from image classification, object detection to generative diffusion model, while also achieving remarkable improvement over various leading binarization algorithms based on different architectures including both CNNs and Transformers.