Small Object Detection Research Articles

Ripe Tomato Detection Algorithm Based on Improved YOLOv9.

Recognizing ripe tomatoes is a crucial aspect of tomato picking. To ensure the accuracy of inspection results, You Only Look Once version 9 (YOLOv9) has been explored as a fruit detection algorithm. To tackle the challenge of identifying tomatoes and the low accuracy of small object detection in complex environments, we propose a ripe tomato recognition algorithm based on an enhanced YOLOv9-C model. After collecting tomato data, we used Mosaic for data augmentation, which improved model robustness and enriched experimental data. Improvements were made to the feature extraction and down-sampling modules, integrating HGBlock and SPD-ADown modules into the YOLOv9 model. These measures resulted in high detection performance with precision and recall rates of 97.2% and 92.3% in horizontal and vertical experimental comparisons, respectively. The module-integrated model improved accuracy and recall by 1.3% and 1.1%, respectively, and also reduced inference time by 1 ms compared to the original model. The inference time of this model was 14.7 ms, which is 16 ms better than the RetinaNet model. This model was tested accurately with mAP@0.5 (%) up to 98%, which is 9.6% higher than RetinaNet. Its increased speed and accuracy make it more suitable for practical applications. Overall, this model provides a reliable technique for recognizing ripe tomatoes during the picking process.

YOLOv8-GDCI: Research on the Phytophthora Blight Detection Method of Different Parts of Chili Based on Improved YOLOv8 Model

Smart farms are crucial in modern agriculture, but current object detection algorithms cannot detect chili Phytophthora blight accurately. To solve this, we introduced the YOLOv8-GDCI model, which can detect the disease on leaves, fruits, and stem bifurcations. The model uses RepGFPN for feature fusion, Dysample upsampling for accuracy, CA attention for feature capture, and Inner-MPDIoU loss for small object detection. In addition, we also created a dataset of chili Phytophthora blight on leaves, fruits, and stem bifurcations, and conducted comparative experiments. The results manifest that the YOLOv8-GDCI model demonstrates outstanding performance across a gamut of comprehensive indicators. In comparison with the YOLOv8n model, the YOLOv8-GDCI model demonstrates an improvement of 0.9% in precision, an increase of 1.8% in recall, and a remarkable enhancement of 1.7% in average precision. Although the FPS decreases slightly, it still exceeds the industry standard for real-time object detection (FPS > 60), thus meeting the requirements for real-time detection.

Scale-adaptive salience supervision and dynamic token filtering for small object detection in remote sensing images

Abstract Recently, DETR-like detectors, which have shown remarkable performance in general object detection, face limitations when dealing with remote sensing images primarily containing small objects. Mainstream two-stage DETR-like models employ a pipeline that selects and processes a small portion of informative tokens, which enhances performance but also shows a high dependency on token selection. The current static token selection strategies lead to inconsistencies between the static selection criteria and dynamic token updates. Additionally, in remote sensing images, the limited information available for small objects and their inherent sensitivity to pixel shifts further degrade detection performance. To address this, we propose Scale-Adaptive Salience DETR (SAS DETR), a two-stage DETR-like method. SAS DETR incorporates dynamic token filtering, which uses a global threshold predictor to determine the token filtering ratio for each layer of the encoder. This approach selects an appropriate filtering ratio for different network layers while maintaining consistency between the foreground confidence map and token updates. Furthermore, we introduce a novel scale-adaptive salience supervision mechanism that adaptively scales the salience computation area based on object size, ensuring the model more effectively supervises small objects and utilizes the information within tokens without compromising the detection performance for objects of other sizes. Finally, we employ Scale-adaptive Intersection over Union to reduce the impact of pixel shifts on small objects. With these improvements, our proposed SAS DETR achieves 25.2% AP on the AI-TOD-V2 dataset with 24 training epochs and 50.4% AP on the COCO 2017 dataset with 12 training epochs.

A Directional Enhanced Adaptive Detection Framework for Small Targets

Due to the challenges posed by limited size and features, positional and noise issues, and dataset imbalance and simplicity, small object detection is one of the most challenging tasks in the field of object detection. Consequently, an increasing number of researchers are focusing on this area. In this paper, we propose a Directional Enhanced Adaptive (DEA) detection framework for small targets. This framework effectively combines the detection accuracy advantages of two-stage methods with the detection speed advantages of one-stage methods. Additionally, we introduce a Multi-Scale Object Adaptive Slicing (MASA) module and an improved IoU-based aggregation module that integrate with this framework to enhance detection performance. For better comparison, we use the F1 score as one of the evaluation metrics. The experimental results demonstrate that our DEA framework improves the performance of various backbone detection networks and achieves better comprehensive detection performance than other proposed methods, even though our network has not been trained on the test dataset while others have.

A YOLO-Based Method for Head Detection in Complex Scenes.

Detecting objects in intricate scenes has always presented a significant challenge in the field of machine vision. Complex scenes typically refer to situations in images or videos where there are numerous densely distributed and mutually occluded objects, making the object detection task even more difficult. This paper introduces a novel head detection algorithm, YOLO-Based Head Detection in Complex Scenes (YOLO-HDCS). Firstly, in complex scenes, head detection typically involves a large number of small objects that are randomly distributed. Traditional object detection algorithms struggle to address the challenge of small object detection. For this purpose, two new modules have been constructed: one is a feature fusion module based on context enhancement with scale adjustment, and the other is an attention-based convolutional module. These modules are characterized by high detection efficiency and high accuracy. They significantly improve the model's multi-scale detection capabilities, thus enhancing the detection ability of the system. Secondly, it was found in practical operations that the original IoU function has a serious problem with overlapping detection in complex scenes. There is an IoU function that can ensure that the final selection boxes cover the object as accurately as possible without overlapping. This not only improves the detection performance but also greatly aids in enhancing the detection efficiency and accuracy. Our method achieves impressive results for head detection in complex scenarios, with average accuracy of 82.2%, and has the advantage of rapid loss convergence during training.

BEV-TinySpotter: A novel BEV perception method considering multi-dimensional feature fusion of small target

The BEV perceptual integrity depends on the ability to accurately capture multi-scale targets. However, the detection of multi-scale targets, especially small targets, is often affected by perceptual noise, information sparsity, and target occlusion, which subsequently affecting the accuracy of BEV perception and causing certain hidden danger to driving safety. To address these issues, we propose a novel BEV perception method BEV-TinySpotter (BEV-TS), which introduced two feature detection enhancement modules and one analytical causal reasoning module to highlighting small object features and improving the accuracy and robustness of small object detection. Specifically, we first enhance the feature recognition of small targets using the multi-scale spatial information acquisition module, and introduce the distribution and migration weights of small targets reflected in the time-series information. Then, we jointly employ global pooling and local pooling to differentiate the acquisition paths of the information to improve the sensitivity of capturing information for small targets. Additionally, we design causal inference graphs to post-fuse the recognition results of BEV, which utilizes causal reasoning to construct a state description mechanism for small targets, further enhancing the accuracy of BEV perception from a logical reasoning perspective. Compared with other SOTA BEV methods, the proposed method effectively enhances the ability to capture and infer the status of small targets during BEV perception, thus significantly improves the completeness of driving environment sensing.

Small Object Detection in UAV Remote Sensing Images Based on Intra-Group Multi-Scale Fusion Attention and Adaptive Weighted Feature Fusion Mechanism

In view of the issues of missed and false detections encountered in small object detection for UAV remote sensing images, and the inadequacy of existing algorithms in terms of complexity and generalization ability, we propose a small object detection model named IA-YOLOv8 in this paper. This model integrates the intra-group multi-scale fusion attention mechanism and the adaptive weighted feature fusion approach. In the feature extraction phase, the model employs a hybrid pooling strategy that combines Avg and Max pooling to replace the single Max pooling operation used in the original SPPF framework. Such modifications enhance the model’s ability to capture the minute features of small objects. In addition, an adaptive feature fusion module is introduced, which is capable of automatically adjusting the weights based on the significance and contribution of features at different scales to improve the detection sensitivity for small objects. Simultaneously, a lightweight intra-group multi-scale fusion attention module is implemented, which aims to effectively mitigate background interference and enhance the saliency of small objects. Experimental results indicate that the proposed IA-YOLOv8 model has a parameter quantity of 10.9 MB, attaining an average precision (mAP) value of 42.1% on the Visdrone2019 test set, an mAP value of 82.3% on the DIOR test set, and an mAP value of 39.8% on the AI-TOD test set. All these results outperform the existing detection algorithms, demonstrating the superior performance of the IA-YOLOv8 model in the task of small object detection for UAV remote sensing.

ATBHC-YOLO: aggregate transformer and bidirectional hybrid convolution for small object detection

Object detection using UAV images is a current research focus in the field of computer vision, with frequent advancements in recent years. However, many methods are ineffective for challenging UAV images that feature uneven object scales, sparse spatial distribution, and dense occlusions. We propose a new algorithm for detecting small objects in UAV images, called ATBHC-YOLO. Firstly, the MS-CET module has been introduced to enhance the model’s focus on global sparse features in the spatial distribution of small objects. Secondly, the BHC-FB module is proposed to address the large-scale variance of small objects and enhance the perception of local features. Finally, a more appropriate loss function, WIoU, is used to penalise the quality variance of small object samples and further enhance the model’s detection accuracy. Comparison experiments on the DIOR and VEDAI datasets validate the effectiveness and robustness of the improved method. By conducting experiments on the publicly available UAV benchmark dataset Visdrone, ATBHC-YOLO outperforms the state-of-the-art method(YOLOv7) by 3.5%.

GreenFruitDetector: Lightweight green fruit detector in orchard environment.

Detecting green fruits presents significant challenges due to their close resemblance in color to the leaves in an orchard environment. We designed GreenFruitDetector, a lightweight model based on an improved YOLO v8 architecture, specifically for green fruit detection. In the Backbone network, we replace ordinary convolution with Deformable Convolution to enhance the extraction of geometric features. Additionally, we designed MCAG-DC (Multi-path Coordinate Attention Guided Deformer Convolution) to replace the convolution in C2f, enhancing the Backbone's feature extraction capability when encountering occlusion problems. For the Neck part of the algorithm, we designed a Fusion-neck structure that integrates spatial detail information from feature maps at different scales, thereby enhancing the network's ability to extract multi-scale information. Additionally, we devised a new detection head that incorporates multi-scale information, significantly improving the detection of small and distant objects. Finally, we applied channel pruning techniques to reduce the model size, parameter count, and FLOPs to 50%, 55%, and 44% of the original, respectively. We trained and evaluated the improved model on three green fruit datasets. The accuracy of the improved model reached 94.5%, 84.4%, and 85.9% on the Korla Pear, Guava, and Green Apple datasets, respectively, representing improvements of 1.17%, 1.1%, and 1.77% over the baseline model. The mAP@0.5 increased by 0.72%, 6.5%, and 0.9%, respectively, and the recall rate increased by 1.97%, 1.1%, and 0.49%, respectively.

Mamba-UAV-SegNet: A Multi-Scale Adaptive Feature Fusion Network for Real-Time Semantic Segmentation of UAV Aerial Imagery

Accurate semantic segmentation of high-resolution images captured by unmanned aerial vehicles (UAVs) is crucial for applications in environmental monitoring, urban planning, and precision agriculture. However, challenges such as class imbalance, small-object detection, and intricate boundary details complicate the analysis of UAV imagery. To address these issues, we propose Mamba-UAV-SegNet, a novel real-time semantic segmentation network specifically designed for UAV images. The network integrates a Multi-Head Mamba Block (MH-Mamba Block) for enhanced multi-scale feature representation, an Adaptive Boundary Enhancement Fusion Module (ABEFM) for improved boundary-aware feature fusion, and an edge-detail auxiliary training branch to capture fine-grained details. The practical utility of our method is demonstrated through its application to farmland segmentation. Extensive experiments on the UAV-City, VDD, and UAVid datasets show that our model outperforms state-of-the-art methods, achieving mean Intersection over Union (mIoU) scores of 71.2%, 77.5%, and 69.3%, respectively. Ablation studies confirm the effectiveness of each component and their combined contributions to overall performance. The proposed method balances segmentation accuracy and computational efficiency, maintaining real-time inference speeds suitable for practical UAV applications.

A Reparameterization Feature Redundancy Extract Network for Unmanned Aerial Vehicles Detection

In unmanned aerial vehicles (UAVs) detection, challenges such as occlusion, complex backgrounds, motion blur, and inference time often lead to false detections and missed detections. General object detection frameworks encounter difficulties in adequately tackling these challenges, leading to substantial information loss during network downsampling, inadequate feature fusion, and being unable to meet real-time requirements. In this paper, we propose a Real-Time Small Object Detection YOLO (RTSOD-YOLO) model to tackle the various challenges faced in UAVs detection. We further enhance the adaptive nature of the Adown module by incorporating an adaptive spatial attention mechanism. This mechanism processes the downsampled feature maps, enabling the model to better focus on key regions. Secondly, to address the issue of insufficient feature fusion, we employ combined serial and parallel triple feature encoding (TFE). This approach fuses scale-sequence features from both shallow features and twice-encoded features, resulting in a new small-scale object detection layer. While enhancing the global context awareness of the existing detection layers, this also enriches the small-scale object detection layer with detailed information. Since rich redundant features often ensure a comprehensive understanding of the input, which is a key characteristic of deep neural networks, we propose a more efficient redundant feature generation module. This module generates more feature maps with fewer parameters. Additionally, we introduce reparameterization techniques to compensate for potential feature loss while further improving the model’s inference speed. Experimental results demonstrate that our proposed RTSOD-YOLO achieves superior detection performance, with mAP50/mAP50:95 reaching 97.3%/51.7%, which represents improvement of 3%/3.5% over YOLOv8, and 2.6%/0.1% higher than YOLOv10. Additionally, it has the lowest parameter count and FLOPs, making it highly efficient in terms of computational resources.

Object Detection and Tracking in Maritime Environments in Case of Person-Overboard Scenarios: An Overview

We examine the current state of the art and the related research on the automated detection and tracking of small objects—or persons—in the context of a person-overboard (POB) scenario and present the associated governing relationship between different technologies, platforms, and approaches as a system of systems. A novel phase model, structuring a POB scenario, comprises three phases: (1) detection, (2) search and track, and (3) rescue. Within these phases, we identify the central areas of responsibility and describe in detail the phases (1) and (2). We emphasize the importance of a high-level representation of different systems and their interactions to comprehensively represent the complexity and dynamics of POB scenarios. Our systematic classification and detailed description of the technologies and methods used provide valuable insights to support future regulatory and research activities. Our primary aim is to advance the development of corresponding technologies and standards.

CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion

CDANet: a small object detection model for unmanned aerial vehicle images based on cross-layer guidance and skip dilated fusion

Vision-guided crack identification and size quantification framework for dam underwater concrete structures

Remotely operated vehicles with cameras provide a non-contact inspection solution for dam underwater structural defect detection. However, manual information extraction methods suffer from problems like labor costs and high misjudgment. This study proposes an integrated dam underwater crack identification and size quantification framework using machine vision and deep learning. First, a real-time automatic detection framework for dam underwater cracks is built via You Only Look Once v5 (YOLOv5), and four different backbone detectors are introduced to balance detection accuracy and speed. Then, the Swin-Transformer module is inserted into the YOLOv5 model to enhance its feature extraction capability and small object detection capability. Next, a method for measuring the true size of cracks was constructed based on deep learning and infrared laser rangefinders. In this study, physical model experiments and actual engineering projects are combined to validate the generalization capability of the proposed crack identification and size quantification method. Experimental results show that the Swin-Transformer-based YOLOv5 model effectively balances detection accuracy and speed with a precision of 0.986, a recall of 0.979, a mean average precision of 0.985, and a frame rate of 68 frames per second in detecting underwater crack images with 768 × 576 pixels. In addition, the proposed method can accurately identify and detect cracks in complex underwater scenes, including obstacle interference, tilt shooting angle, uneven illumination, and turbid water scenarios. Moreover, the method proposed in this paper can quantify the overall size and geometric parameters of underwater cracks with relatively small errors.

YOLO-ResTinyECG: ECG-based lightweight embedded AI arrhythmia small object detector with pruning methods

ObjectiveThis study presents a YOLO-ResTinyECG, a novel model for small object detection in electrocardiogram (ECG) images, emphasizing longer time-window lengths for improving the throughput of arrhythmia detection. Methods: The proposed ResTinyECG backbone consists of four-stage Res-blocks (2/4/3/2) with input ECG image sizes of 320 × 320 × 1 pixels and 160 × 160 × 1 pixels, with a significant parameter reduction of around 20 %/95 %/74 % compared to existing backbones like Shufflenet-v2/CSPDarknet53-tiny and the latest YOLOv7-tiny structure. Moreover, this study applied magnitude-based filter pruning and dependency graph (DepGraph) pruning for parameter optimization. The ECG images undergo processing with time-window lengths of 5/10/15/20 s, accompanied by min–max normalization, and weighted loss is applied for class balance during learning, and non-maximum suppression with edge removal is used to compute and select the final output bounding boxes of ECG heartbeats. Experiments: Experiments conducted on the PhysioNet MIT-BIH arrhythmia ECG database focus on nine classes, including normal (N), atrial premature beat (A), ventricular premature beat (V), left bundle branch block (L), right bundle branch block (R), fusion of ventricular beat (FVN), paced beat (P), fusion of paced and normal beat (FPN), and others (remaining heartbeats). Results: The proposed ResTinyECG-320 exhibits impressive mean Average Precision (mAP) scores of 94.76 %/94.85 %/93.96 %/87.12 % in 6-class detection and 92.35 %/91.96 %/90.58 %/83.34 % in 9-class detection across different time-window lengths. The model demonstrates only a marginal 7 %∼9% decrease in performance when utilizing longer time-window lengths. After applying magnitude-based filter pruning, ResTinyECG can be reduced to a minimal 0.1 million parameters. Furthermore, ResTinyECG-160 achieves a competitive mAP of 93.92 % in 6-class detection and a faster processing speed of 100.6 ECG segments per second (SPS) on a PC compared to Shufflenet-v2 (77.3 SPS). In conclusion, YOLO-ResTinyECG surpasses existing backbones, exhibiting superior mAP in both 6/9-class detection scenarios, and its deployment on an embedded AI platform validates its real-time detection capabilities.

Efficient Small Object Detection You Only Look Once: A Small Object Detection Algorithm for Aerial Images.

Aerial images have distinct characteristics, such as varying target scales, complex backgrounds, severe occlusion, small targets, and dense distribution. As a result, object detection in aerial images faces challenges like difficulty in extracting small target information and poor integration of spatial and semantic data. Moreover, existing object detection algorithms have a large number of parameters, posing a challenge for deployment on drones with limited hardware resources. We propose an efficient small-object YOLO detection model (ESOD-YOLO) based on YOLOv8n for Unmanned Aerial Vehicle (UAV) object detection. Firstly, we propose that the Reparameterized Multi-scale Inverted Blocks (RepNIBMS) module is implemented to replace the C2f module of the Yolov8n backbone extraction network to enhance the information extraction capability of small objects. Secondly, a cross-level multi-scale feature fusion structure, wave feature pyramid network (WFPN), is designed to enhance the model's capacity to integrate spatial and semantic information. Meanwhile, a small-object detection head is incorporated to augment the model's ability to identify small objects. Finally, a tri-focal loss function is proposed to address the issue of imbalanced samples in aerial images in a straightforward and effective manner. In the VisDrone2019 test set, when the input size is uniformly 640 × 640 pixels, the parameters of ESOD-YOLO are 4.46 M, and the average mean accuracy of detection reaches 29.3%, which is 3.6% higher than the baseline method YOLOv8n. Compared with other detection methods, it also achieves higher detection accuracy with lower parameters.

Dynamic feature and context enhancement network for faster detection of small objects

While traditional object detection methods have achieved significant success in recent years, their performance in detecting small objects in aerial remote sensing images remains unsatisfactory. Small objects often occupy only a few pixels, leading to a loss of fine-grained information. This paper proposes a dynamic feature and context enhancement network (DFCE) to address noise and pixel-level region weighting issues in small object detection. The DFCE network effectively detects small objects by dynamically selecting features within regions and establishing connections between local and global contextual information. The introduced dynamic multi-dimensional attention (DMA) module selects key information via a crossover mechanism and assigns different weights to highlight important features. Based on DMA, the regional feature processing (RFP) module and multi-dimensional pool transformer (MPT) module are developed to capture key information and contextual information, respectively. Experimental results demonstrate that the DFCE network improves average precision (AP) by 3.1%, 9%, and 2.2% on two remote sensing datasets and one conventional dataset, achieving an inference speed of 30 frames per second (FPS). Given these advancements, the DFCE model’s powerful key feature extraction and contextual association capabilities show strong potential for broader applications, including sign language recognition, defect detection in industrial settings, and more.

Small Object Detection with Small Samples Using High-Resolution Remote Sensing Images

Abstract Interpretation of remote sensing images has become a research hotspot in the field of remote sensing in recent years. It is currently widely applied in areas such as mapping, dynamic monitoring, earth resource surveys and geological disaster investigation. Compared to traditional methods, remote sensing image target detection and recognition methods based on deep learning have achieved significant improvements in accuracy. However, these methods often face challenges such as sample scarcity, interference from complex background, limited feature information, and the dependence on discriminative key feature regions for recognizing fine-grained targets. Addressing these challenges, this paper conducts research on small target detection methods using high-resolution remote sensing images. It explores deep learning theories and methods such as feature enhancement and attention mechanisms within a supervised learning framework. The proposed target detection model consists of four parts: Deep feature extraction module, which extracts features of small targets at multiple scales. Feature enhancement module, which enhances the feature differences between the background and small targets at different scales. Target detection module based on enhanced features. Loss function for optimizing network parameters. Experimental validation shows that this model can effectively extract feature information of small targets under sample-scarce conditions, achieving outstanding results in small target detection in remote sensing images.

Research on YOLOv3 Method Based on SE Module

Object detection is a crucial and challenging task in computer vision. With advancements in deep learning technology, YOLOv3 has become a widely adopted and efficient object detection algorithm. However, YOLOv3 encounters challenges when managing complex scenes and detecting small objects. To tackle these challenges, this research introduces an enhanced YOLOv3 architecture incorporating the Squeeze-and-Excitation (SE) module to improve feature representation capabilities. The SE module captures the interdependencies between channels and dynamically adjusts their feature responses, enhancing the models representation capability. By integrating the SE module into YOLOv3, this study seeks to substantially improve the models effectiveness in complex scenes and small object detection. Experimental results indicate that the enhanced YOLOv3 surpasses the original model on the COCO2017 dataset, validating the effectiveness of this method. Additionally, the improved architecture further enhances detection accuracy and robustness while maintaining efficient detection speed. The contribution of this study lies in proposing an effective feature enhancement method, introducing innovative concepts and techniques for object detection.

DeployFusion: A Deployable Monocular 3D Object Detection with Multi-Sensor Information Fusion in BEV for Edge Devices.

To address the challenges of suboptimal remote detection and significant computational burden in existing multi-sensor information fusion 3D object detection methods, a novel approach based on Bird's-Eye View (BEV) is proposed. This method utilizes an enhanced lightweight EdgeNeXt feature extraction network, incorporating residual branches to address network degradation caused by the excessive depth of STDA encoding blocks. Meantime, deformable convolution is used to expand the receptive field and reduce computational complexity. The feature fusion module constructs a two-stage fusion network to optimize the fusion and alignment of multi-sensor features. This network aligns image features to supplement environmental information with point cloud features, thereby obtaining the final BEV features. Additionally, a Transformer decoder that emphasizes global spatial cues is employed to process the BEV feature sequence, enabling precise detection of distant small objects. Experimental results demonstrate that this method surpasses the baseline network, with improvements of 4.5% in the NuScenes detection score and 5.5% in average precision for detection objects. Finally, the model is converted and accelerated using TensorRT tools for deployment on mobile devices, achieving an inference time of 138 ms per frame on the Jetson Orin NX embedded platform, thus enabling real-time 3D object detection.