A Full Featured Configurable Accelerator for Object Detection With YOLO

Pestana, Daniel; Miranda, Pedro R.; Lopes, João D.; Duarte, Rodrigo; Véstias, Mário P.; Neto, Horácio C.; Sousa, J.T. de

doi:10.1109/access.2021.3081818

Cited by 53 publications

(22 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, by the up-sampling of the detection head of YOLOv5 in the addition of a detection head based on the Path Aggregation Network, the challenge of dense object distribution was mitigated. Authors [40] in their work demonstrated that the YOLOV5 architecture is not only a fast object detector but also achieved accuracy that is plausible to that of Faster R-CNN. In a similar study, authors [41] proved that the YOLO model can adequately classify and detect multiple objects in real-time, which makes it suitable as an underlying detector for the anti-drone system.…”

Section: Related Work On Drone Detection Techniques and Technologiesmentioning

confidence: 84%

DRONET: Multi-Tasking Framework for Real-Time Industrial Facility Aerial Surveillance and Safety

Ajakwe

Ihekoronye

Kim

et al. 2022

Drones

View full text Add to dashboard Cite

The security of key and critical infrastructures is crucial for uninterrupted industrial process flow needed in strategic management as these facilities are major targets of invaders. The emergence of non-military use of drones especially for logistics comes with the challenge of redefining the anti-drone approach in determining a drone’s harmful status in the airspace based on certain metrics before countering it. In this work, a vision-based multi-tasking anti-drone framework is proposed to detect drones, identifies the airborne objects, determines its harmful status through perceived threat analysis, and checks its proximity in real-time prior to taking an action. The model is validated using manually generated 5460 drone samples from six (6) drone models under sunny, cloudy, and evening scenarios and 1709 airborne objects samples of seven (7) classes under different environments, scenarios (blur, scales, low illumination), and heights. The proposed model was compared with seven (7) other object detection models in terms of accuracy, sensitivity, F1-score, latency, throughput, reliability, and efficiency. The simulation result reveals that, overall, the proposed model achieved superior multi-drone detection accuracy of 99.6%, attached object identification of sensitivity of 99.80%, and F1-score of 99.69%, with minimal error, low latency, and less computational complexity needed for effective industrial facility aerial surveillance. A benchmark dataset is also provided for subsequent performance evaluation of other object detection models.

show abstract

Section: Related Work On Drone Detection Techniques and Technologiesmentioning

confidence: 84%

DRONET: Multi-Tasking Framework for Real-Time Industrial Facility Aerial Surveillance and Safety

Ajakwe

Ihekoronye

Kim

et al. 2022

Drones

View full text Add to dashboard Cite

show abstract

“…Compared to [40], the proposed work was slower and less efficient in terms of BRAMs but more efficient in terms of LUT and similarly efficient in terms DSP. The lower BRAM efficiency has to do with the fact that BRAM contents (weights and activations) are shared by multiple cores.…”

Section: Comparison With Other Fpga Implementationsmentioning

confidence: 97%

“…The solution does not apply to real-time applications but provides a YOLO solution in a low-cost FPGA. Recently, another implementation of Tiny-YOLOv3 [40] with a 16-bit fixed-point format achieved 32 FPS in a UltraScale XCKU040 FPGA. The accelerator runs the CNN and pre-and post-processing tasks with the same architecture.…”

Section: Convolutional Neural Network Accelerators In Fpgamentioning

confidence: 99%

“…The work proposed in this article modifies the architecture designed in [40] so that it can be mapped to low-density FPGAs. New addressing mechanisms to support feature map tilling, optimization of cores to improve DSP utilization, and multiple DMAs to take advantage of multiple memory controller ports were the main hardware modifications.…”

Section: Convolutional Neural Network Accelerators In Fpgamentioning

confidence: 99%

See 1 more Smart Citation

Configurable Hardware Core for IoT Object Detection

et al. 2021

Self Cite

View full text Add to dashboard Cite

Object detection is an important task for many applications, like transportation, security, and medical applications. Many of these applications are needed on edge devices to make local decisions. Therefore, it is necessary to provide low-cost, fast solutions for object detection. This work proposes a configurable hardware core on a field-programmable gate array (FPGA) for object detection. The configurability of the core allows its deployment on target devices with diverse hardware resources. The object detection accelerator is based on YOLO, for its good accuracy at moderate computational complexity. The solution was applied to the design of a core to accelerate the Tiny-YOLOv3, based on a CNN developed for constrained environments. However, it can be applied to other YOLO versions. The core was integrated into a full system-on-chip solution and tested with the COCO dataset. It achieved a performance from 7 to 14 FPS in a low-cost ZYNQ7020 FPGA, depending on the quantization, with an accuracy reduction from 2.1 to 1.4 points of mAP50.

show abstract

“…Also, FPGA can quickly update the design with its programmability and can be deployed quickly. Therefore, there are several researches on DCNN accelerator based on FPGA [9,10].…”

Section: Introductionmentioning

confidence: 99%

A digital signal processor‐efficient accelerator for depthwise separable convolution

Huang

Liu

et al. 2022

Electronics Letters

View full text Add to dashboard Cite

Recent researches on deep convolution neural networks have proposed some compact networks, such as MobileNet, but its main computation, depthwise separable convolution (DWC), which reduces the reusable data and improves the requirement of data loading efficiency. Although DWC can effectively reduce the amount of network computation, it needs a special accelerator to enhance the inference speed. This paper proposes a high-performance accelerator for DWC based on the commonly used acceleration platform field-programmable gate array. The proposed accelerator supports the computation of both standard convolutions (SCs) and DWC as well as two activation functions. In addition, two data storage formats are used to maintain the data loading efficiency for different input requirements of SC and DWC under high parallelism. Furthermore, a processing unit that can execute two 8 × 8bit multiplications inside one digital signal processor (DSP) is designed to make the best use of the DSP hardware resources. Finally, the accelerator is implemented on ZYNQ ZC706 at 200 MHz. Consuming only 392 DSPs, the accelerator achieves 134.5 giga operations per second (GOPS) and 209.4 frames per second (FPS) on MobileNet V1 as well as 96.4 GOPS and 250.4 FPS on MobileNet V2. Experimental results demonstrate that this design provides a better DSP efficiency than previous works.

show abstract

A Full Featured Configurable Accelerator for Object Detection With YOLO

Cited by 53 publications

References 33 publications

DRONET: Multi-Tasking Framework for Real-Time Industrial Facility Aerial Surveillance and Safety

DRONET: Multi-Tasking Framework for Real-Time Industrial Facility Aerial Surveillance and Safety

Configurable Hardware Core for IoT Object Detection

A digital signal processor‐efficient accelerator for depthwise separable convolution

Contact Info

Product

Resources

About