Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation

Zheng, Zhaohui; Wang, Ping; Ren, Dongwei; Liu, Wei; Ye, Rongguang; Hu, Qinghua; Zuo, Wangmeng

doi:10.1109/tcyb.2021.3095305

Cited by 701 publications

(251 citation statements)

References 50 publications

Supporting

Mentioning

251

Contrasting

Order By: Relevance

“…Furthermore, it can measure the spatial as well as the morphological discrepancy of two RBBox when they are non-overlapping, as shown in Figure 4 (c). This idea has frequently appeared in the IoU family of loss functions, such as GIoU (Rezatofighi et al 2019), DIoU (Zheng et al 2020), and CIoU (Zheng et al 2021).…”

Section: Continuous and Differentiablementioning

confidence: 99%

Polygon-to-Polygon Distance Loss for Rotated Object Detection

Yang

Chen

Zhong

et al. 2022

AAAI

View full text Add to dashboard Cite

There are two key issues that limit further improvements in the performance of existing rotational detectors: 1) Periodic sudden change of the parameters in the rotating bounding box (RBBox) definition causes a numerical discontinuity in the loss (such as smoothL1 loss). 2) There is a gap of optimization asynchrony between the loss in the RBBox regression and evaluation metrics. In this paper, we define a new distance formulation between two convex polygons describing the overlapping degree and non-overlapping degree. Based on this smooth distance, we propose a loss called Polygon-to-Polygon distance loss (P2P Loss). The distance is derived from the area sum of triangles specified by the vertexes of one polygon and the edges of the other. Therefore, the P2P Loss is continuous, differentiable, and inherently free from any RBBox definition. Our P2P Loss is not only consistent with the detection metrics but also able to measure how far, as well as how similar, a RBBox is from another one even when they are completely non-overlapping. These features allow the RetinaNet using the P2P Loss to achieve 79.15% mAP on the DOTA dataset, which is quite competitive compared with many state-of-the-art rotated object detectors.

show abstract

Section: Continuous and Differentiablementioning

confidence: 99%

Polygon-to-Polygon Distance Loss for Rotated Object Detection

Yang

Chen

Zhong

et al. 2022

AAAI

View full text Add to dashboard Cite

show abstract

“…We can cite RetinaNet [39], which is designed to focus on the most challenging samples (e.g., small objects) by multiplying a term proportional to the network's confidence into the classical cross-entropy loss. Other methods modify the standard IoU loss, including Intersection over Detection, Generalized IoU [141], Wasserstein distance [142], and Complete IoU [143]; The detailed explanation of these methods can be found in Section 6.2.1.…”

Section: Loss Function Regularizationmentioning

confidence: 99%

A Guide to Image and Video based Small Object Detection using Deep Learning : Case Study of Maritime Surveillance

Rekavandi¹,

Liu²,

Boussaid³

et al. 2022

Preprint

View full text Add to dashboard Cite

Small object detection (SOD) in optical images and videos is a challenging problem that even state-of-the-art generic object detection methods fail to accurately localize and identify such objects. Typically, small objects appear in real-world due to large camera-object distance. Because small objects occupy only a small area in the input image (e.g., less than 10%), the information extracted from such a small area is not always rich enough to support decision making. Multidisciplinary strategies are being developed by researchers working at the interface of deep learning and computer vision to enhance the performance of SOD deep learning based methods. In this paper, we provide a comprehensive review of over 160 research papers published between 2017 and 2022 in order to survey this growing subject. This paper summarizes the existing literature and provide a taxonomy that illustrates the broad picture of current research. We investigate how to improve the performance of small object detection in maritime environments, where increasing performance is critical. By establishing a connection between generic and maritime SOD research, future directions have been identified. In addition, the popular datasets that have been used for SOD for generic and maritime applications are discussed, and also well-known evaluation metrics for the state-of-the-art methods on some of the datasets are provided.

show abstract

“…Here, ˆi x , ˆi y , ˆi w , and ˆi h correspond to the central coordinates and width and height of the actual box, respectively, In this paper, Complete-IoU [29] is used for bounding box regression, which could take into account the distance, scale, overlap area, and aspect ratio between target and anchor compared with IoU. Therefore, regression becomes more stable and effectively avoids divergence of the training process model.…”

Section: Loss Function Improvementmentioning

confidence: 99%

Tomato Maturity Classification Based on SE-YOLOv3-MobileNetV1 Network under Nature Greenhouse Environment

Zhao

Wang

et al. 2022

Agronomy

View full text Add to dashboard Cite

The maturity level of tomato is a key factor of tomato picking, which directly determines the transportation distance, storage time, and market freshness of postharvest tomato. In view of the lack of studies on tomato maturity classification under nature greenhouse environment, this paper proposes a SE-YOLOv3-MobileNetV1 network to classify four kinds of tomato maturity. The proposed maturity classification model is improved in terms of speed and accuracy: (1) Speed: Depthwise separable convolution is used. (2) Accuracy: Mosaic data augmentation, K-means clustering algorithm, and the Squeeze-and-Excitation attention mechanism module are used. To verify the detection performance, the proposed model is compared with the current mainstream models, such as YOLOv3, YOLOv3-MobileNetV1, and YOLOv5 in terms of accuracy and speed. The SE-YOLOv3-MobileNetV1 model is able to distinguish tomatoes in four kinds of maturity, the mean average precision value of tomato reaches 97.5%. The detection speed of the proposed model is 278.6 and 236.8 ms faster than the YOLOv3 and YOLOv5 model. In addition, the proposed model is considerably lighter than YOLOv3 and YOLOv5, which meets the need of embedded development, and provides a reference for tomato maturity classification of tomato harvesting robot.

show abstract

Enhancing Geometric Factors in Model Learning and Inference for Object Detection and Instance Segmentation

Cited by 701 publications

References 50 publications

Polygon-to-Polygon Distance Loss for Rotated Object Detection

Polygon-to-Polygon Distance Loss for Rotated Object Detection

A Guide to Image and Video based Small Object Detection using Deep Learning : Case Study of Maritime Surveillance

Tomato Maturity Classification Based on SE-YOLOv3-MobileNetV1 Network under Nature Greenhouse Environment

Contact Info

Product

Resources

About