SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution Dynamic convolution achieves better performance for efficient CNNs at the cost of negligible FLOPs increase However, the performance increase can not match the significantly expanded number of parameters, which is the main bottleneck in real-world applications
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution We propose the Sparse Dynamic Convolution (SD-Conv) to improve the parameter eficiency of dynamic convolution by marrying the dynamic convolution and sparsity to maintain the advantage of both worlds
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution In this paper, we propose a new framework, Sparse Dynamic Convolution (SD-CONV), to naturally integrate these two paths such that it can inherit the advantage of dynamic mechanism and sparsity
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution This paper designs a binary mask derived from a learnable threshold to prune static kernels, significantly reducing the parameters and computational cost but achieving higher performance in Imagenet-1K and hopes the SD-Conv could be an efficient alternative to conventional dynamic convolutions
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution In this paper, we study the trade-off between accuracy and speed when building an object detection system based on convolutional neural networks We consider three main families of detectors ---
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution Abstract: Dynamic convolution achieves better performance for efficient CNNs at the cost of negligible FLOPs increase However, the performance increase can not match the significantly expanded number of parameters, which is the main bottleneck in real-world applications
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution Article "SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution" Detailed information of the J-GLOBAL is an information service managed by the Japan Science and Technology Agency (hereinafter referred to as "JST")
Chenbo Jiang Sparse Dynamic Convolution (SD-Conv) naturally integrate sparse convolution and dynamic convolution such that it can inherit the advantage of dynamic mechanism and sparsity
SD-Conv: Towards the Parameter-Efficiency of Dynamic Convolution We propose an efficient training method for CNN compression via dynamic parameter rank pruning Our experiments show that the proposed method can yield substantial storage savings while maintaining or even enhancing classification performance