An improved small target detection method based on Yolo V3

Zhang, Gongguo; Wei, Jingsong

doi:10.1109/ecie52353.2021.00054

Cited by 11 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The backbone and head of a convolutional neural network are the two fundamental components of the YOLOv8 architecture, which is an improvement over earlier iterations of the YOLO algorithm [8]. A revised Currently, CS architecture that consists of thirty-five convolutional layers and uses cross-stage fractional connections to enhance the transfer of data between layers serves as the foundation of YOLOv8.…”

Section: Models Architecturementioning

confidence: 99%

Improved small-object detection using YOLOv8: A comparative study

Huang,

Wang,

Xiao

et al. 2024

ACE

View full text Add to dashboard Cite

In the last decade or so, deep neural networks have evolved at a rapid pace, where computer vision has been constantly refreshing its best performance and has been integrated into our lives. In the field of target detection, YOLO model is a popular real-time target detection algorithm model that is fast, efficient, and accurate. This research aims to optimize the latest YOLOv8 model to improve its detection of small objects and compare it with another different version of YOLO models. To achieve this goal, we used the classical deep learning algorithm YOLOv8 as a benchmark and made several improvements and optimizations. We optimized the definition of the detection head, narrowed its perceptual field, and increased its number, allowing the model to better focus on the detailed information of small objects. We compared the optimized YOLOv8 model with other classical YOLO models, including YOLOv3 and YOLOv5n. The experimental results show that our optimized model improves small object detection with higher accuracy. This research provides an effective solution for small object detection with good application prospects. With the continuous development and improvement of the technology, we believe that the YOLO algorithm will continue to play an essential role in object detection and provide a reliable solution for various real-time applications.

show abstract

Section: Models Architecturementioning

confidence: 99%

Improved small-object detection using YOLOv8: A comparative study

Huang,

Wang,

Xiao

et al. 2024

ACE

View full text Add to dashboard Cite

show abstract

“…Song et al [12] achieved pose estimation for multi-target cows with minimal or low occlusion using the partial affinity field algorithm. Zhang et al [13] implemented beef cattle pose estimation using an improved YOLOv3 algorithm, capable of recognizing standing and lying behaviors based on extracted skeleton information.Furthermore, researchers such as Lin [14] and Qi [15] employed HRNet, a highly regarded network in human pose estimation, for the pose estimation of birds and giant pandas, respectively. Sun et al [16] leveraged skeleton sequence information of monkeys for behavior recognition, introducing a skeleton behavior recognition method based on a global spatio-temporal coding network, which demonstrated superior accuracy and reduced computational parameters compared to the ST-GCN model in the context of monkey behavior recognition tasks.…”

Section: Introduction and Related Workmentioning

confidence: 99%

A method for estimating pig pose in a breeding scenario

Qin,

Mao,

Wang

et al. 2024

International Conference on Optics and Machine Vision (ICOMV 2024)

View full text Add to dashboard Cite

Aiming at the problem of low automation in pig farms, this paper proposes a new pig posture estimation method based on breeding scenarios for intelligent monitoring of pig farms.Firstly, the video image data of indoor and outdoor scenes in pig breeding scenarios were collected and labeled, and a self-constructed pig posture estimation dataset was built; Secondly,Resnet50, VGG16 and MobileNetV2 were used as the back-bone network, and the three methods based on coordinate regression, heat map and simple coordinate classi-fication were analyzed experimentally, and the Simple Coordinate Classification (SimCC) algorithm with the optimal effect was selected as the extraction method of key points of pigs;Finally, we integrated High Resolution Network(HRNet)and HRFormer, which incorporates Transformer modules, as backbone net-works. They were combined with the SimCC to formulate an effective pig pose estimation framework. The experimental results show that the mAP of HRFormer-SimCC reaches 83.2%, which is an average im-provement of 7.2% over the use of traditional CNN model and 0.4% over the HRNet-SimCC, and the float-ing-point computation and parameter counts of HRFormer-SimCC are only 45.05% and 36.48% of it. This is more suitable to be deployed in breeding environments and provides a theoretical basis for intelligent moni-toring of pig farms.

show abstract

“…Introduced efficient spatio-temporal interaction (ESI) module, and the C3 module 24 was replaced in Backbone 24 with ESI module. The shallow bottleneck (SB) module was used to retain more shallow semantic information 25 , thus solved the problem of large loss of small target information 26 . We added a fourth detector head to the original network for the output of minimal targets.…”

Section: Introductionmentioning

confidence: 99%

GBS-YOLOv5: Improved Algorithm for UAV Intelligent Transportation Based on YOLOv5

Liu

Duan

Lou

et al. 2023

Preprint

View full text Add to dashboard Cite

As the road traffic situation becomes complex, the task of traffic management takes on an increasingly heavy load. The air-to-ground traffic administration network of drones has become an important tool to promote the high quality of traffic police work in many places. Drones can be used instead of a large number of human beings to perform daily tasks, as: traffic offense detection, daily crowd detection, etc. UAVs are less accurate at detecting drones because they operate from the air and take pictures of small targets. To address the problem of low accuracy of UAVs in detecting small targets, we proposed the GBS-YOLOv5 model. It was an improvement on the original YOLOv5 model. The default YOLOv5 model suffered from severe loss of small target information and under-utilization of shallow feature information as the depth of the feature extraction network deepened. We deepened the feature extraction network by introducing an efficient spatiotemporal interaction (ESI) module. We added the spatial pyramid convolution (SPC) module as a detection head for tiny-sized targets. A shallow bottleneck (SB) was introduced to better preserve the detailed information of small targets in the shallow features. The introduction of recursive gated convolution in the feature fusion section enabled better interaction of higher-order spatial semantic information. The GBS-YOLOv5 algorithm conducted experiments showing that the value of mAP@0.5 was 35.3% and the mAP@0.5:0.95 was 20.0%. Compared to the default YOLOv5 algorithm was boosted by 4.0% and 3.5% respectively.

show abstract

An improved small target detection method based on Yolo V3

Cited by 11 publications

References 7 publications

Improved small-object detection using YOLOv8: A comparative study

Improved small-object detection using YOLOv8: A comparative study

A method for estimating pig pose in a breeding scenario

GBS-YOLOv5: Improved Algorithm for UAV Intelligent Transportation Based on YOLOv5

Contact Info

Product

Resources

About