Search papers, labs, and topics across Lattice.
This paper introduces Smart Scissor, a novel framework that combines dynamic image cropping with a compound shrinking strategy to optimize convolutional neural networks (CNNs) for embedded hardware. By accurately cropping foreground objects from images, the framework reduces spatial redundancy while maintaining critical features, leading to improved recognition accuracy even at lower resolutions. Experimental results on ImageNet-1K show that Smart Scissor reduces the computational cost of ResNet50 by 41.5% while enhancing top-1 accuracy by 0.3%, outperforming existing methods like HRank by 4.1% in accuracy at equivalent computational costs.
Smart Scissor reduces CNN computational costs by over 40% while actually improving accuracy, challenging the notion that efficiency comes at the expense of performance.
Scaling down the resolution of input images can greatly reduce the computational overhead of convolutional neural networks (CNNs), which is promising for edge AI. However, as an image usually contains much spatial redundancy, e.g., background pixels, directly shrinking the whole image will lose important features of the foreground object and lead to severe accuracy degradation. In this paper, we propose a dynamic image cropping framework to reduce the spatial redundancy by accurately cropping the foreground object from images. To achieve the instance-aware fine cropping, we introduce a lightweight foreground predictor to efficiently localize and crop the foreground of an image. The finely cropped images can be correctly recognized even at a small resolution. Meanwhile, computational redundancy also exists in CNN architectures. To pursue higher execution efficiency on resource-constrained embedded devices, we also propose a compound shrinking strategy to coordinately compress the three dimensions (depth, width, resolution) of CNNs. Eventually, we seamlessly combine the proposed dynamic image cropping and compound shrinking into a unified compression framework, Smart Scissor, which is expected to significantly reduce the computational overhead of CNNs while still maintaining high accuracy. Experiments on ImageNet-1K demonstrate that our method reduces the computational cost of ResNet50 by 41.5% while improving the top-1 accuracy by 0.3%. Moreover, compared to HRank, the state-of-theart CNN compression framework, our method achieves 4.1% higher top-1 accuracy at the same computational cost. The codes and data are available at https://github.com/ntuliuteam/smart-scissor
