Conventional main memory can no longer meet the requirements of low energy consumption and massive data storage in an artificial intelligence Internet of Things (AIoT) system. Moreover, the efficiency is decreased due to the swapping of data between the main memory and storage. This paper presents a hybrid storage class memory system to reduce the energy consumption and optimize IO performance. Phase change memory (PCM) brings the advantages of low static power and a large capacity to a hybrid memory system. In order to avoid the impact of poor write performance in PCM, a migration scheme implemented in the memory controller is proposed. By counting the write times and row buffer miss times in PCM simultaneously, the write-intensive data can be selected and migrated from PCM to dynamic random-access memory (DRAM) efficiently, which improves the performance of hybrid storage class memory. In addition, a fast mode with a tmpfs-based, in-memory file system is applied to hybrid storage class memory to reduce the number of data movements between memory and external storage. Experimental results show that the proposed system can reduce energy consumption by 46.2% on average compared with the traditional DRAM-only system. The fast mode increases the IO performance of the system by more than 30 times compared with the common ext3 file system.
Convolutional neural networks have been widely applied in the field of computer vision. In convolutional neural networks, convolution operations account for more than 90% of the total computational workload. The current mainstream approach to achieving high energy-efficient convolution operations is through dedicated hardware accelerators. Convolution operations involve a significant amount of weights and input feature data. Due to limited on-chip cache space in accelerators, there is a significant amount of off-chip DRAM memory access involved in the computation process. The latency of DRAM access is 20 times higher than that of SRAM, and the energy consumption of DRAM access is 100 times higher than that of multiply–accumulate (MAC) units. It is evident that the “memory wall” and “power wall” issues in neural network computation remain challenging. This paper presents the design of a hardware accelerator for convolutional neural networks. It employs a dataflow optimization strategy based on on-chip data reordering. This strategy improves on-chip data utilization and reduces the frequency of data exchanges between on-chip cache and off-chip DRAM. The experimental results indicate that compared to the accelerator without this strategy, it can reduce data exchange frequency by up to 82.9%.
1 October 2025 in IEICE Transactions on Information and Systems
Object detection from a UAV perspective faces challenges such as object occlusion, unclear boundaries, and small target sizes, resulting in reduced detection accuracy. Additionally, traditional object detection algorithms have large parameter counts, making them unsuitable for resource-constrained edge devices. To address this issue, we propose a lightweight small-object detection algorithm: Sky-YOLO. Specifically, we introduce the MSFConv multi-scale feature map fusion convolution module into the backbone network to enhance feature extraction capability. The Neck part is replaced with the L-BiFPN module to reduce parameter count and strengthen feature fusion between layers. Additionally, based on the characteristics of UAV imagery, we incorporate the WIOU loss function, enabling efficient detection of blurred and occluded targets. Experimental results show that the Sky-YOLO model, with 60% fewer parameters than the original model, still achieves 39.7% accuracy on the VisDrone2019 validation dataset, a 6.7% improvement in accuracy over the original model.
Deep learning significantly advances object detection. Post process, a critical component of this process, selects valid bounding boxes to represent true targets during inference and assigns boxes and labels to these objects during training to optimize the loss function. However, post process constitutes a substantial portion of the total processing time for a single image. This inefficiency primarily arises from the extensive Intersection over Union (IoU) calculations required between numerous redundant bounding boxes in post-processing algorithms. To reduce the redundant IoU calculations, we introduce a classification prioritization strategy during both training and inference post processes. Additionally, post process involves sorting operations that contribute to inefficiency. To minimize unnecessary comparisons in Top-K sorting, we have improved the bitonic sorter by developing a hybrid bitonic algorithm. These improvements have effectively accelerated post process. Given the similarities between training and inference post processes, we unify four typical post-processing algorithms and design a hardware accelerator based on this framework. Our accelerator achieves at least 7.55 times the speed in inference post process compared to recent accelerators. When compared to the RTX 2080 Ti system, our proposed accelerator offers at least 21.93 times the speed for training post process and 19.89 times for inference post process, thereby significantly enhancing the efficiency of loss function minimization.
For neural network accelerators with General Matrix Multiplication (GEMM) as the computational core, the input feature maps of convolution must be converted into 2D matrices through the Im2col operation. Conventional approaches utilize CPUs to execute Im2col management and data transfer operations. Conventional methods suffer from memory expansion due to redundant data in overlapping convolutional windows, thus incurring non-negligible memory access energy consumption and transmission latency overheads. This severely limits the feasibility of efficient GEMM acceleration in resource-constrained edge devices. This paper proposes a novel Low Memory Access Im2col Method (LMAI2C) and present its dedicated hardware implementation. By restructuring data from overlapping convolutional windows, LMAI2C significantly reduces DRAM memory access volume while improving feature map transfer efficiency. Experimental results on convolutional layers of the YOLOv4-tiny network demonstrate that LMAI2C reduces DDR memory access by approximately 79.8% compared to traditional methods. Furthermore, LMAI2C demonstrates an average speedup of 69 times compared to CPU-based methodologies and 43 times over DMA-accelerated CPU implementations.