Medical image segmentation is an essential process that facilitates the precise extraction and localization of diseased areas from medical pictures. It can provide clear quantifiable information to support clinicians in making final decisions. However, due lack explicit modeling global relationships CNNs, they are unable fully use long-range dependencies among several locations. In this paper, we propose a novel model extract local semantic features images by utilizing CNN visual transformer...

Edge artificial intelligence applications impose rigorous demands on local hardware to improve throughput and energy efficiency. Computing-in-memory (CIM) architectures provide high parallel energy-efficient solutions accelerate the multiply-and-accumulate (MAC) operations in neural networks (NNs). While SRAM-based charge-domain CIM is achieving thousands of TOPS/W efficiency, it encounters limitations when dealing with full NN model deployments where both activations weights are signed....

Traffic sign detection plays a pivotal role in autonomous driving systems. The intricacy of the model necessitates high-performance hardware. Real-world traffic environments exhibit considerable variability and diversity, posing challenges for effective feature extraction by model. Therefore, it is imperative to develop that not only highly accurate but also lightweight. In this paper, we proposed YOLO-ADual, novel lightweight Our method leverages C3Dual Adown modules as replacements CPS CBL...

Traffic sign detection plays a pivotal role in autonomous driving systems. The intricacy of the model necessitates high-performance hardware. Real-world traffic environments exhibit considerable variability and diversity, posing challenges for effective feature extraction by model. Therefore, it is imperative to develop that not only highly accurate but also lightweight. In this paper, we proposed YOLO-ADual, novel lightweight Our method leverages C3Dual Adown modules as replacements CPS CBL...

Flying bird detection has recently attracted increasing attention in computer vision. However, compared to conventional object tasks, it is much more challenging trap flying birds infrared videos due small target size, complex backgrounds, and dim shapes. In order solve the problem of poor performance caused by insufficient feature information birds, this paper suggests a method detecting outdoor environments using image pre-processing deep learning, called temporal Variation filtering (TVF)...

While the performance of deep convolutional neural networks for image super-resolution (SR) has improved significantly, rapid increase memory and computation requirements hinders their deployment on resource-constrained devices. Quantized networks, especially binary (BNN) SR have been proposed to significantly improve model inference efficiency but suffer from large degradation. We observe activation distribution demonstrates very pixel-to-pixel, channel-to-channel, image-to-image variation,...

Stochastic computing (SC) has emerged as a promising paradigm for neural acceleration. However, how to accelerate the state-of-the-art Vision Transformer (ViT) with SC remains unclear. Unlike convolutional networks, ViTs introduce notable compatibility and efficiency challenges because of their nonlinear functions, e.g., softmax Gaussian Error Linear Units (GELU). In this paper, first time, ViT accelerator based on end-to-end SC, dubbed ASCEND, is proposed. ASCEND co-designs circuits...

Depthwise separable neural network models with fewer parameters, such as MobileNet, are more friendly to edge AI devices. They replace the standard convolution depthwise convolution, which consists of a (DW) and pointwise (PW) convolution. Most prior computing-in-memory (CIM) works [1–5] only optimize multiply-and-accumulate (MAC) operations for one these two types. Thus, when performing recent SRAM-based CIMs still face limitations in energy efficiency, throughput, memory utilization (Fig....

Traditional diesel-based airport service vehicles are characterized by a heavy-duty, high-usage-frequency nature and high carbon intensity per vehicle hour. Transforming these into electric would reduce CO2 emissions potentially save energy costs in the context of rising fuel prices, if proper management (ASEVs) is performed. To perform such an management, this paper proposes new customized rollout approach, as near-optimal control method for ASEV dynamics model, which models states, their...

With the rapid advancements of deep learning in recent years, hardware accelerators are continuously deployed more and safety-critical applications such as autonomous driving robotics. While usually fabricated with advanced technology nodes for high performance energy efficiency, they also prone to timing errors under process, voltage, temperature, aging (PVTA) variations. By revisiting physical sources errors, we show that most accelerator caused by a specific subset input patterns, defined...

With the rapid advancements of deep learning in recent years, hardware accelerators are continuously deployed more and safety-critical applications such as autonomous driving robotics. While usually fabricated with advanced technology nodes for high performance energy efficiency, they also prone to timing errors under process, voltage, temperature, aging (PVTA) variations. By revisiting physical sources errors, we show that most accelerator caused by a specific subset input patterns, defined...

The Nanjing Circumvallation has been severely damaged due to natural factors, human interference, and urban expansion. Restoring this edifice serves not only as a preservation of historikcal cultural values, but also an advancement towards sustainable development. This research proposes environmental ethical decision-making model (EEDM) grounded in development indicators. is then applied the restoration Circumvallation. Based on feedback from trial section, renovation segment, guided by...