Critically Compressed Quantized Convolution Neural Network based High Frame Rate and Ultra-Low Delay Fruit External Defects Detection

Zhang, Jihan; Huang, Dongmei; Hu, Tingting; Fuchikami, Ryuji; Ikenaga, Takeshi

doi:10.23919/mva51890.2021.9511388

Cited by 2 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To improve the processing speed of fruit detection, some researchers have proposed using lightweight network structures. Zhang et al [26] proposed bit depth degression quantization for defect classification. Xu et al [27] proposed an improved YOLOv3-tiny method that used depth-wise separable CNN, achieving an inferential speed of 40.35 ms. Lawal [14] modified the YOLOv3 model to use densely architecture incorporation, spatial pyramid pooling, and Mish function activation, achieving a detection time of 44 ms. Chen et al [28] combined MobileNet V2 and YOLO, and reached a fast detection speed of 17 ms for citrus sorting in the conveyor.…”

Section: Fruit and Vegetable Defect Detectionmentioning

confidence: 99%

Global to multi‐scale local architecture with hardwired CNN for 1‐ms tomato defect detection

Li,

Hu,

Fuchikami

et al. 2024

IET Image Processing

Self Cite

View full text Add to dashboard Cite

A 1 millisecond (1‐ms) vision system that guarantees high efficiency and timely response for tomato defect detection is essential for factory automation. Because of various defect appearances, recently many existing researches focus on CNN based defect detection, but few of them attempt to reach high processing speed to adapt to the factorial assembly line. This paper proposes a global to multi‐scale local based parallel architecture with hardwired CNN for tomato defect detection. This architecture breaks down image‐wise detection into pixel‐wise localization and block‐wise classification. The pixel‐wise localization utilizes tomato‐aware information as constraints for localization performance. The block‐wise classification uses a fully pipelined network structure to obtain the classification result for each block as the pixel stream moves through the network. The classification network has a six‐layer lightweight network structure with quantization for hardwired type implementation on FPGA. The experiment results show that the proposed architecture processes 1000 FPS images with 0.9476 ms/frame delay. And for detection performance, this architecture keeps at 80.18%, only 1.31% lower than ResNet50 based detection system.

show abstract

Section: Fruit and Vegetable Defect Detectionmentioning

confidence: 99%

Global to multi‐scale local architecture with hardwired CNN for 1‐ms tomato defect detection

Li,

Hu,

Fuchikami

et al. 2024

IET Image Processing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Alignment: The object may not be aligned with the sensor during an inspection, making it challenging to acquire accurate images or sensor data. Researchers have proposed various methods to overcome these problems, such as using multiple High-Frame Rate Camera Sensors [18] and advanced deep-learning architectures [19].…”

Section: Introductionmentioning

confidence: 99%

Optimization of Computational Resources for Real-Time Product Quality Assessment Using Deep Learning and Multiple High Frame Rate Camera Sensors

Wibowo

Setiawan

Afrisal

et al. 2023

ASI

View full text Add to dashboard Cite

Human eyes generally perform product defect inspection in Indonesian industrial production lines; resulting in low efficiency and a high margin of error due to eye tiredness. Automated quality assessment systems for mass production can utilize deep learning connected to cameras for more efficient defect detection. However, employing deep learning on multiple high frame rate cameras (HFRC) causes the need for much computation and decreases deep learning performance, especially in the real-time inspection of moving objects. This paper proposes optimizing computational resources for real-time product quality assessment on moving cylindrical shell objects using deep learning with multiple HFRC Sensors. Two application frameworks embedded with several deep learning models were compared and tested to produce robust and powerful applications to assess the quality of production results on rotating objects. Based on the experiment results using three HFRC Sensors, a web-based application with tensorflow.js framework outperformed desktop applications in computation. Moreover, MobileNet v1 delivers the highest performance compared to other models. This result reveals an opportunity for a web-based application as a lightweight framework for quality assessment using multiple HFRC and deep learning.

show abstract

Critically Compressed Quantized Convolution Neural Network based High Frame Rate and Ultra-Low Delay Fruit External Defects Detection

Cited by 2 publications

References 10 publications

Global to multi‐scale local architecture with hardwired CNN for 1‐ms tomato defect detection

Global to multi‐scale local architecture with hardwired CNN for 1‐ms tomato defect detection

Optimization of Computational Resources for Real-Time Product Quality Assessment Using Deep Learning and Multiple High Frame Rate Camera Sensors

Contact Info

Product

Resources

About