CASSOD-Net: Cascaded and Separable Structures of Dilated Convolution for Embedded Vision Systems and Applications
This work addresses performance bottlenecks for embedded vision systems by improving efficiency and accuracy, though it is incremental as it builds on existing dilated convolution methods.
The paper tackled the inefficiency of dilated convolutions in neural networks by proposing CASSOD-Net, a module that replaces traditional dilated filters with cascaded 2x2 filters, achieving higher accuracy in face detection with only 47% of filter weights and accelerating computations by 2.78 times in hardware.
The field of view (FOV) of convolutional neural networks is highly related to the accuracy of inference. Dilated convolutions are known as an effective solution to the problems which require large FOVs. However, for general-purpose hardware or dedicated hardware, it usually takes extra time to handle dilated convolutions compared with standard convolutions. In this paper, we propose a network module, Cascaded and Separable Structure of Dilated (CASSOD) Convolution, and a special hardware system to handle the CASSOD networks efficiently. A CASSOD-Net includes multiple cascaded $2 \times 2$ dilated filters, which can be used to replace the traditional $3 \times 3$ dilated filters without decreasing the accuracy of inference. Two example applications, face detection and image segmentation, are tested with dilated convolutions and the proposed CASSOD modules. The new network for face detection achieves higher accuracy than the previous work with only 47% of filter weights in the dilated convolution layers of the context module. Moreover, the proposed hardware system can accelerate the computations of dilated convolutions, and it is 2.78 times faster than traditional hardware systems when the filter size is $3 \times 3$.