角点和前景区域 IoU 损失:在界限框回归中更好地定位小物体
Delong Cai1,2, Zhaoyun Zhang1, Zhi Zhang1
1School of Electrical Engineering and Intelligentization, DongGuan University of Technology, Dongguan 523000, China.
Sensors (Basel, Switzerland)
|July 11, 2023
概括
新的角点和前景区域 IoU (CFIoU) 损失通过使用角点距离和目标前景区域来改善小物体检测. 这种方法提高了本地化准确度和融合速度,以获得更好的物体检测性能.
科学领域:
- 计算机视觉 计算机视觉
- 机器学习 机器学习
- 深度学习 (Deep Learning) 是一种深度学习.
背景情况:
- 界限框回归对于对象检测准确性至关重要,特别是对于小物体.
- 现有的欧盟 (BIoU) 损失的广泛交叉面临着缓慢的融合和目标框附近无效的合适信息的挑战.
- 当前的局部化损失函数通常不充分利用目标前景区域的空间信息.
研究的目的:
- 提出一种新的角点和前景区域 IoU (CFIoU) 损失函数,以解决对象检测的界限框回归的局限性.
- 增强空间目标数据的合适信息和利用,特别是用于小物体检测.
- 提高对象检测模型中的收速度和定位精度.
主要方法:
- 引入了CFIoU损失,通过将规范的中心点距离替换为规范的角点距离,以改善盒子接近时的装配.
- 将适应性目标前景区域信息集成到损失函数中,以提供更丰富的上下文数据以进行优化.
- 通过模拟实验和量化比较与YOLOv5,YOLOv8和SSD算法的主流BIoU损失验证了CFIoU损失.
主要成果:
- 对于YOLOv5s和YOLOv8s的VisDrone2019和SODA-D数据集,CFIoU损失显著改善了性能,在回忆和mAP指标中显示出显著的收益.
- 在VisDrone2019上,使用CFIoU的YOLOv5看到+3.12%的召回和+2.73%的mAP@0.5;YOLOv8看到+1.72%的召回和+0.60%的mAP@0.5.
- 在SODA-D上,YOLOv5s与CFIoU实现了+6%的召回和+13.08%的mAP@0.5;YOLOv8s看到+3.36%的召回和+3.66%的mAP@0.5.
- 结合CFIoU损失的SSD算法显示了AP (+5.59%) 和AP75 (+5.37%) 的改进,即使对于未专门用于小物体检测的算法也显示了有效性.
结论:
- 拟议的CFIoU损失函数对于小物体检测任务是有效和优越的.
- CFIoU损失提高了本地化准确性和趋同性,优于现有的BIoU损失.
- CFIoU的损失通过提高不同物体检测架构的性能来证明多功能性,包括那些在小物体检测方面不那么熟练的架构.
相关概念视频
01:18Reducing Line Loss
174
In a three-phase circuit, line loss is an indicator of energy dissipated as heat due to the resistance of transmission lines. To address this, incorporating transformers into the system—a step-up transformer at the source and a step-down transformer at the load—is a strategic solution. Two three-phase transformers are introduced to improve this.
With a step-up transformer at the source, the voltage is increased, thereby reducing the current in the transmission lines since power loss...
With a step-up transformer at the source, the voltage is increased, thereby reducing the current in the transmission lines since power loss...
174
02:07Improving Translational Accuracy
11.6K
Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...
11.6K
01:11Residuals and Least-Squares Property
7.4K
The vertical distance between the actual value of y and the estimated value of y. In other words, it measures the vertical distance between the actual data point and the predicted point on the line
If the observed data point lies above the line, the residual is positive, and the line underestimates the actual data value for y. If the observed data point lies below the line, the residual is negative, and the line overestimates the actual data value for y.
The process of fitting the best-fit...
If the observed data point lies above the line, the residual is positive, and the line underestimates the actual data value for y. If the observed data point lies below the line, the residual is negative, and the line overestimates the actual data value for y.
The process of fitting the best-fit...
7.4K
01:10Line Loss
273
The different configurations of source-load connections include wye (star) and delta connections. The relationship between line and phase voltages and currents varies depending on the configuration. When the source is supplying power, it is transmitted through the wires to the load, and during this transmission, some power is absorbed by the wires, leading to line loss.
Line loss impacts power delivery efficiency in a balanced three-phase circuit. The symmetry in such a circuit simplifies the...
Line loss impacts power delivery efficiency in a balanced three-phase circuit. The symmetry in such a circuit simplifies the...
273
01:27Margin of Error
4.4K
The margin of error is also called the maximum error of an estimate. The margin of error is the maximum possible or expected difference between the observed sample parameter value and the actual population parameter value. For proportion, it is the maximum difference between the value of sample proportion obtained from the data and the true value of population proportion. As the true value of the population parameter is not known, the margin of error is calculated using the sample statistic.
4.4K
01:15Distance Corrections
51
To achieve precise distance measurements, especially in surveying and construction, certain corrections must be applied to account for potential sources of error like the standardization errors, temperature variations, and slope adjustments.Standardization error emerges when measurement equipment undergoes changes, such as wear, repairs, or weather impacts. To address this, surveyors compare the equipment’s readings to a standard. This process identifies any deviation that might lead to...
51