Search

The accuracy of current deep learning algorithms has certainly increased. However, deploying deep learning networks on edge devices with limited resources is challenging due to their inherent depth and high parameter count. Here, we proposed an improved YOLO model based on an attention mechanism and receptive field (RFA-YOLO) model, applying the MobileNeXt network as the backbone to reduce parameters and complexity, adopting the Receptive Field Block (RFB) and Efficient Channel Attention (ECA) modules to improve the detection accuracy of multi-scale and small objects. Meanwhile, an FPGA-based model deployment solution was proposed to implement parallel acceleration and low-power deployment of the detection algorithm model, which achieved real-time object detection for optical remote sensing images. We implement the proposed DPU and Vitis AI-based object detection algorithms with FPGA deployment to achieve low power consumption and real-time performance requirements. Experimental results on DIOR dataset demonstrate the effectiveness and superiority of our RFA-YOLO model for object detection algorithms. Moreover, to evaluate the performance of the proposed hardware implementation, it was implemented on a Xilinx ZCU104 board. Results of the experiments for hardware and software simulation show that our DPU-based hardware implementation are more power efficient than central processing units (CPUs) and graphics processing units (GPUs), and have the potential to be applied to onboard processing systems with limited resources and power consumption. [ABSTRACT FROM AUTHOR]

With the development of big data, analyzing the environmental benefits of transportation systems by artificial intelligence has become a hot issue in recent years. The ground traffic changes can be overlooked from a high-altitude perspective, using the technology of multi-temporal remote sensing change detection. We proposed a novel unsupervised algorithm by combining the image transformation and deep learning method. The new algorithm for remote sensing images is named multi-attention slow feature analysis (ASFA). In this model, three parts perform different functions respectively. The first part records to the K-BoVW to classify the categories of the ground objects as a channel parameter. The second part is a residual convolution with multiple attention mechanisms including temporal, spatial, and channel attention. Feature extraction and updating are completed at this link. Finally, we put the updated features in the slow feature analysis to highlight the variant components which we want and then generate the change map visually. Experiments on three very high-resolution datasets verified that the ASFA has a better performance than four basic change detection algorithms and an improved SFA algorithm. More importantly, this model works well for traffic road detection and helps us analyze the environmental benefits of traffic changes. [ABSTRACT FROM AUTHOR]

In recent years, deep convolutional neural networks (DCNNs) have shown their predominant place in building extraction. Generally, the performance of DCNNs depends on the quality of parameter learning. To improve the performance, many approaches chose to enlarge the network size or employ additional plug-in modules. However, there exists large redundancy in parameter learning of neural networks. We argue that the methods totally relying on parameter learning cannot fully exploit the useful information of the given data, and feature learned by such methods is not comprehensive enough. In this paper, we propose an efficient memory module, named EMM , to enhance the learning ability of DCNNs in building extraction. In our method, we encode each batch of data into memory key-value pairs and store them in memory banks, in which each memory will dynamically update during the training process. Inspired by the non-local module, we propose to use a tri-branch way for a more efficient memory retrieval process. By training with our memory module, DCNNs can learn more holistic and discriminative features efficiently and effectively. We conduct intensive experiments on three challenging public datasets. Compared with popular segmentation networks on all three datasets, our method can achieve competitive performance while keeping relatively fewer parameters. • Using deep neural networks to extracts building footprints automatically. • Integrated memory module enables network learning to have more distinguishing feature. • The proposed memory module consumes less computation than the original module. • Implicit relation between features is established to capture discriminant features. [ABSTRACT FROM AUTHOR]