Deep learning has been constantly improving in recent years and a significant
number of researchers have devoted themselves to the research of defect
detection algorithms. Detection and recognition of small and complex targets is
still a problem that needs to be solved. The authors of this research would
like to present an improved defect detection model for detecting small and
complex defect targets in steel surfaces. During steel strip production
mechanical forces and environmental factors cause surface defects of the steel
strip. Therefore the detection of such defects is key to the production of
high-quality products. Moreover surface defects of the steel strip cause great
economic losses to the high-tech industry. So far few studies have explored
methods of identifying the defects and most of the currently available
algorithms are not sufficiently effective. Therefore this study presents an
improved real-time metallic surface defect detection model based on You Only
Look Once (YOLOv5) specially designed for small networks. For the smaller
features of the target the conventional part is replaced with a depth-wise
convolution and channel shuffle mechanism. Then assigning weights to Feature
Pyramid Networks (FPN) output features and fusing them increases feature
propagation and the networks characterization ability. The experimental results
reveal that the improved proposed model outperforms other comparable models in
terms of accuracy and detection time. The precision of the proposed model
achieved by @mAP is 77.5% on the Northeastern University Dataset NEU-DET and
70.18% on the GC10-DET datasets

基于 YOLOv5 的改进型缺陷检测模型在钢表面上检测小且复杂的缺陷目标，在精度和检测时间方面表现优于其他可比较的模型。

利用通道洗牌的深度学习加速金属表面缺陷检测

Faster Metallic Surface Defect Detection Using Deep Learning with  Channel Shuffling

The width of a neural network matters since increasing the width will
necessarily increase the model capacity. However, the performance of a network
does not improve linearly with the width and soon gets saturated. In this case,
we argue that increasing the number of networks (ensemble) can achieve better
accuracy-efficiency trade-offs than purely increasing the width. To prove it,
one large network is divided into several small ones regarding its parameters
and regularization components. Each of these small networks has a fraction of
the original one's parameters. We then train these small networks together and
make them see various views of the same data to increase their diversity.
During this co-training process, networks can also learn from each other. As a
result, small networks can achieve better ensemble performance than the large
one with few or no extra parameters or FLOPs, \ie, achieving better
accuracy-efficiency trade-offs. Small networks can also achieve faster
inference speed than the large one by concurrent running. All of the above
shows that the number of networks is a new dimension of model scaling. We
validate our argument with 8 different neural architectures on common
benchmarks through extensive experiments. The code is available at
https://github.com/FreeformRobotics/Divide-and-Co-training.

通过将大型神经网络分解成多个小型网络，并通过共同训练和学习来提高网络的多样性，小型网络可以在不增加参数和 FLOPs 的情况下实现更好的集成性能和更快的推理速度。我们在八个不同的神经结构和常见基准测试中验证了该论点，代码可在 https://github.com/FreeformRobotics/Divide-and-Co-training 中获得。