D-CenterNet: An Anchor-Free Detector With Knowledge Distillation for Industrial Defect Detection

Liu, Zhiwei; Lyu, Wentao; Wang, Chengqun; Guo, Qing; Zhou, Di; Xu, Weiqiang

doi:10.1109/tim.2022.3204332

Cited by 13 publications

(7 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, the integration of the CBAM attention mechanism with a feature fusion module enables the model to selectively focus on pertinent feature channels and spatial locations, significantly boosting the discriminative power of the feature representation and thereby increasing overall accuracy. KD-LightNet [14] introduced an efficient and lightweight defect detection network optimized for edge computing scenarios. The network architecture, LightNet, is crafted using structure reparameterization to boost feature extraction capabilities while reducing model complexity.…”

Section: Related Workmentioning

confidence: 99%

Towards High Quality PCB Defect Detection Leveraging State-of-the-Art Hybrid Models

Nguyen,

Nguyen

2024

IJACSA

View full text Add to dashboard Cite

The automatic detection of defects in printed circuit boards (PCBs) is a critical step in ensuring the reliability of electronic devices. This paper introduces a novel approach for PCB defect detection. It incorporates a state-of-the-art hybrid architecture that leverages both convolutional neural networks (CNNs) and transformer-based models. Our model comprises three main components: a Backbone for feature extraction, a Neck for feature map refinement, and a Head for defect prediction. The Backbone utilizes ResNet and Bottleneck Transformer blocks, which are proficient at highlighting small defect features and overcoming the shortcomings of previous models. The Neck module, designed with Ghost Convolution, refines feature maps. It reduces computational demands while preserving the quality of feature representation. This module also facilitates the integration of multi-scale features, essential for accurately detecting a wide range of defect sizes. The Head employs a Fully Convolutional One-stage detection approach, allowing for the prediction process to proceed without reliance on predefined anchors. Within the Head, we incorporate the Wise-IoU loss to refine bounding box regression. This optimizes the model's focus on high-overlap regions and mitigates the influence of outlier samples. Comprehensive experiments on standard PCB datasets validate the effectiveness of our proposed method. The results show significant improvements over existing techniques, particularly in the detection of small and subtle defects.

show abstract

Section: Related Workmentioning

confidence: 99%

Towards High Quality PCB Defect Detection Leveraging State-of-the-Art Hybrid Models

Nguyen,

Nguyen

2024

IJACSA

View full text Add to dashboard Cite

show abstract

“…where p is the center point of the detection frame , , is the theoretical centroid mapping to the exact location region of the feature map.However, in most cases the coordinates of p are floating point, so we adjust the coordinates p as follows:p .So p is rounded down, thus forming an error. The O in the formula indicates the offset output feature map, which is the area where the key points actually fall into [5].…”

Section: ) Networkmentioning

confidence: 99%

Learning CenterNet model for wild animal detection

Wu,

Chang,

et al. 2023

Second International Conference on Electronic Information Technology (EIT 2023)

View full text Add to dashboard Cite

The detection of animal species as a kind of target detection, the application of artificial intelligence techniques to the field of image processing helps us to study animals more intelligently. In this paper, we use CenterNet-based algorithm to achieve target detection of animals. firstly, we perform data enhancement on the used dataset to achieve multi-scene coverage. secondly, we use CenterNet algorithm to train the model on the dataset to improve the accuracy of recognition. finally, we deploy it to the camera for animal recognition in the actual target scene. The experimental results are analyzed and the trained animal detection system recognizes an average accuracy of 90%-97%. This experiment demonstrates the utility of this research in animal detection in the field of image processing.

show abstract

“…To increase the effectiveness of picture feature extraction in complicated backgrounds, Liu et al. [16] suggested a new anchor‐free method based on knowledge distillation and introduced module strip pooling. Han et al.…”

Section: Introductionmentioning

confidence: 99%

“…In recent years, deep learning-based methods have become increasingly popular in the fields of defect recognition [13][14][15][16][17], fault diagnosis [8], industrial process soft sensing [9][10][11], and so on. Deep learning algorithms are used for defect identification in a variety of tasks, such as classification, location, detection, segmentation, and others.…”

Section: Introductionmentioning

confidence: 99%

“…Jiao et al [15] used multi-scale characteristics and context information to remove interference from complicated backgrounds and related objects in bolt defect images. To increase the effectiveness of picture feature extraction in complicated backgrounds, Liu et al [16] suggested a new anchor-free method based on knowledge distillation and introduced module strip pooling. Han et al [17] introduced a new two-stage network for detecting strip defects by differentiating between foreground and background using edge characteristics.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MACN: A cascade defect detection for complex background based on mixture attention mechanism

Zhao,

Wu,

Yuan

et al. 2024

IET Image Processing

View full text Add to dashboard Cite

Defect detection in complex background is a critical issue. To address this issue, this paper proposes the mixture attention mechanism cascade network, in which the new channel attention network is linked with the spatial attention network to create an effective mixed attention network that takes advantage of their respective advantages, adaptively suppresses background noise, and highlights defect features. To ensure the efficiency and effectiveness of effective mixed attention network, the new channel attention network splices the output features of the global average pooling layer and the global maximum pooling layer and then sends the spliced features into a shared network, which is a one‐dimensional convolutional network, and uses cross‐channel interaction for fusion. Furthermore, in order to provide more discriminative feature representation, the authors extract the intermediate features of the region proposal network and input them into effective mixed attention network. Finally, the cascade head is used to refine the predicted bounding box to achieve high‐quality defect location. To demonstrate the superiority and usefulness of this method, it is compared to the latest method using widely used PCB and NEU data sets. A large number of trials demonstrate that this strategy outperforms other methods for detecting defects in complicated backgrounds.

show abstract

D-CenterNet: An Anchor-Free Detector With Knowledge Distillation for Industrial Defect Detection

Cited by 13 publications

References 36 publications

Towards High Quality PCB Defect Detection Leveraging State-of-the-Art Hybrid Models

Towards High Quality PCB Defect Detection Leveraging State-of-the-Art Hybrid Models

Learning CenterNet model for wild animal detection

MACN: A cascade defect detection for complex background based on mixture attention mechanism

Contact Info

Product

Resources

About