Target Detection Task Research Articles

A Review of YOLO-Based Target Detection Methods

Target detection is now a popular topic in computer vision. With the development and iteration of technology, deep learning is constantly emerging. The integration of deep learning in the target detection task has led to rapid improvement in accuracy and speed, among which the You Only Look Once (YOLO) series of methods have the most rapid and varied improvements and upgrades, which have been widely used in the fields of navigation, video surveillance, face detection, text detection, and aerospace, etc. This paper initially provides an overview of the research context, importance, and challenges associated with this domain, compares and analyzes the network structure and implementation of the single-phase target detection method represented by the YOLO series with the two-phase and other improved algorithms, and then introduces the research progress of the target detection algorithms of deep learning, the characteristics of the commonly used datasets, and the key parameters of the evaluation of performance indicators, and then presents a compilation of experimental outcomes associated with several widely recognized algorithms applied to prominent datasets. Subsequently, it enumerates the experimental findings of diverse algorithms on these established datasets. Ultimately, this paper anticipates future research trajectories and developmental trends pertaining to target detection algorithms.

Embedded YOLO v8: Real-time detection of sugarcane nodes in complex natural environments by rapid structural pruning method

In the exploring of sugarcane intelligent harvesting technology, the target detection task of sugarcane node faces serious challenges on low detection accuracy affected by complex natural environment and the time-consuming task to lightweight algorithm structure. In addition, the classical target detection model, such as YOLO (You Only Look Once), with requirement of huge computationally ability puts huge computing pressure to the embedded device for sugarcane nodes detection. In order to solve these problems, this study proposed an improved YOLOv8n-ghost, which adopted Ghost module to build lightweight network and reduce model redundancy while ensuring performance. The dataset was extended and diversified by data classification and enhancement to increasing the robustness and generalization ability of the model. The structured pruning method based on DepGraph (Dependency Graph) was adopted to compress the optimized model, which greatly reduced the complexity of the model. To reduce experimentation time and computational cost, a rapid method for the optimal pruning rate was proposed by combination of big-small step size to quickly achieve the optimal pruning rate with shorter experiment time and satisfied accuracy. By comparison of different pruning methods and lightweight methods, the AP (Average Precision) of model Pruned 65 %-YOLOv8n-ghost was > 90 % with only 0.53 M parameters and 2.2G FLOPs. Finally, the model was accelerated with TensorRT and tested in the embedded device. The real-time speed of the improved model approached 30 frames per second, which was 200 % higher than before and met the requirements of real-time detection of sugarcane node.

Side-Scan Sonar Image Generation Under Zero and Few Samples for Underwater Target Detection

The acquisition of side-scan sonar (SSS) images is complex, expensive, and time-consuming, making it difficult and sometimes impossible to obtain rich image data. Therefore, we propose a novel image generation algorithm to solve the problem of insufficient training datasets for SSS-based target detection. For zero-sample detection, we proposed a two-step style transfer approach. The ray tracing method was first used to obtain an optically rendered image of the target. Subsequently, UA-CycleGAN, which combines U-net, soft attention, and HSV loss, was proposed for generating high-quality SSS images. A one-stage image-generation approach was proposed for few-sample detection. The proposed ADA-StyleGAN3 incorporates an adaptive discriminator augmentation strategy into StyleGAN3 to solve the overfitting problem of the generative adversarial network caused by insufficient training data. ADA-StyleGAN3 generated high-quality and diverse SSS images. In simulation experiments, the proposed image-generation algorithm was evaluated subjectively and objectively. We also compared the proposed algorithm with other classical methods to demonstrate its advantages. In addition, we applied the generated images to a downstream target detection task, and the detection results further demonstrated the effectiveness of the image generation algorithm. Finally, the generalizability of the proposed algorithm was verified using a public dataset.

SGW-YOLOv8n: An Improved YOLOv8n-Based Model for Apple Detection and Segmentation in Complex Orchard Environments

This study addresses the problem of detecting occluded apples in complex unstructured environments in orchards and proposes an apple detection and segmentation model based on improved YOLOv8n-SGW-YOLOv8n. The model improves apple detection and segmentation by combining the SPD-Conv convolution module, the GAM global attention mechanism, and the Wise-IoU loss function, which enhances the accuracy and robustness. The SPD-Conv module preserves fine-grained features in the image by converting spatial information into channel information, which is particularly suitable for small target detection. The GAM global attention mechanism enhances the recognition of occluded targets by strengthening the feature representation of channel and spatial dimensions. The Wise-IoU loss function further optimises the regression accuracy of the target frame. Finally, the pre-prepared dataset is used for model training and validation. The results show that the SGW-YOLOv8n model significantly improves relative to the original YOLOv8n in target detection and instance segmentation tasks, especially in occlusion scenes. The model improves the detection mAP to 75.9% and the segmentation mAP to 75.7% and maintains a processing speed of 44.37 FPS, which can meet the real-time requirements, providing effective technical support for the detection and segmentation of fruits in complex unstructured environments for fruit harvesting robots.

A Knowledge Base Driven Task-Oriented Image Semantic Communication Scheme

With the development of artificial intelligence and computer hardware, semantic communication has been attracting great interest. As an emerging communication paradigm, semantic communication can reduce the requirement for channel bandwidth by extracting semantic information. This is an effective method that can be applied to image acquisition of unmanned aerial vehicles, which can transmit high-data-volume images within the constraints of limited available bandwidth. However, the existing semantic communication schemes fail to adequately incorporate the guidance of task requirements into the semantic communication process and are difficult to adapt to the dynamic changes of tasks. A task-oriented image semantic communication scheme driven by knowledge base is proposed, aiming at achieving high compression ratio and high quality image reconstruction, and effectively solving the bandwidth limitation. This scheme segments the input image into several semantic information unit under the guidance of task requirements by Yolo-World and Segment Anything Model. The assigned bandwidth for each unit is according to the task relevance scores, which enables high-quality transmission of task-related information with lower communication overheads. An improved metric weighted learned perceptual image patch similarity (LPIPS) is proposed to evaluate the transmission accuracy of the novel scheme. Experimental results show that our scheme achieves a notable performance improvement on weighted LPIPS while the same compression ratio compared with traditional image compression schemes. Our scheme has a higher target capture ratio than traditional image compression schemes under the task of target detection.

Research on Traffic Marking Segmentation Detection Algorithm Based on Feature Fusion

Identifying road markings is a very important part of the vehicle environment sensing system and plays a crucial role in a vehicle’s correct understanding of a current traffic situation. However, road traffic markings are interfered with by a variety of factors, such as being obscured and the viewpoint of the vehicle sensors, resulting in large errors in the existing detection methods. In order to make the target detection task applicable to irregular objects or to detection tasks with higher accuracy requirements while reducing the waste of computational resources, this paper improves the accuracy of traffic marking segmentation detection by designing a multi-type traffic marking segmentation detection model based on image segmentation algorithms and designing a segmentation guidance matrix module based on a rank guidance matrix computation method. By constructing a comprehensive traffic marking detection model, a unified road traffic marking detection is achieved. Finally, the new traffic marking datasets ApolloScape-Precise and ApolloScape-Large are constructed based on the existing ApolloScape dataset, and experimental validation is carried out on these two datasets. The results show that the index MIoU (Mean Intersection over Union) of traffic marking segmentation detection reaches 61.44% and 70.15%, thus achieving a more perfect road traffic marking detection and right-of-way information perception and proving the effectiveness of the integrated traffic marking detection method designed in this paper.

Rib Fracture Detection Model Based on Faster-RCNN-SE-FA Algorithm

To address the problem of missed diagnosis in rib fracture detection from CT scans, this study introduces an enhanced model, called Faster-RCNN-SE-FA, which is built upon the traditional Faster-RCNN architecture. The proposed model integrates a novel filter anchor method and thoroughly considers the specific imaging characteristics of ribs in CT images. The preprocessing of the image is followed by applying the Squeeze-and-Excitation (SE) module, which enhances the discrimination of features in the channel dimension while preserving the location Sensitivity (Sen) important for target detection tasks. Consequently, this modification leads to a significant improvement in model performance. Empirical experiments, conducted on CT sequences of 130 rib feature cases provided by the First Affiliated Hospital of Ningbo University, demonstrate that the Faster-RCNN-SE-FA model achieves better Sen and accuracy compared to traditional methods, including the baseline Faster-RCNN.

Dynamic Attending in percussionists and wind players

Percussionists have been shown to have superior temporal perception, but the underlying mechanisms for this advantage deserve further investigation. Dynamic Attending Theory, which posits the way attention aligns with rhythmic temporal structure, may explain why percussionists exhibit superior temporal perception. Percussionists may entrain more strongly to periodic stimuli than other musicians, and possibly at multiple hierarchical levels as opposed to only the tactus. To test these hypotheses, we asked a total of 41 musicians, 20 percussionists, and 21 wind players to complete a target detection task. Participants listened to periodic bursts of noise ( priming bursts) and tried to detect target bursts of noise at different temporal positions, some in phase and some out of phase with the initial noises. Target-burst amplitudes were calibrated to be at each participant’s perceptual threshold. We analyzed the hit rates at the different temporal positions, as well as the Fourier transform of those hit rate curves. No discernible difference between groups was found in the time domain, but in the frequency domain the percussionist group showed stronger fluctuation at a frequency that matched the priming bursts, but not at other hierarchical levels (specifically, double that frequency). Our findings help explain percussionists’ perceptual advantages, although questions about entrainment at different hierarchical levels deserve further consideration.

ESFuse: Weak Edge Structure Perception Network for Infrared and Visible Image Fusion

Infrared and visible image fusion (IVIF) fully integrates the complementary features of different modal images, and the fused image provides a more comprehensive and objective interpretation of the scene compared to each source image, thus attracting extensive attention in the field of computer vision in recent years. However, current fusion methods usually center their attention on the extraction of prominent features, falling short of adequately safeguarding subtle and diminutive structures. To address this problem, we propose an end-to-end unsupervised IVIF method (ESFuse), which effectively enhances fine edges and small structures. In particular, we introduce a two-branch head interpreter to extract features from source images of different modalities. Subsequently, these features are fed into the edge refinement module with the detail injection module (DIM) to obtain the edge detection results of the source image, improving the network’s ability to capture and retain complex details as well as global information. Finally, we implemented a multiscale feature reconstruction module to obtain the final fusion results by combining the output of the DIM with the output of the head interpreter. Extensive IVIF fusion experiments on existing publicly available datasets show that the proposed ESFuse outperforms the state-of-the-art(SOTA) methods in both subjective vision and objective evaluation, and our fusion results perform well in semantic segmentation, target detection, pose estimation and depth estimation tasks. The source code has been availabled.

Road object detection method based on improved YOLOv5 algorithm

Aiming at the existing network's poor recognition of distant targets in road traffic scenes, insufficient expression of target features, and inaccurate target positioning, a road target detection method based on the improved YOLOv5 algorithm is proposed. Firstly, the feature extraction structure of the YOLOv5 algorithm is summarized, and the shortcomings of the original network structure are analyzed; secondly, a small target detection layer is added on the basis of the original network, and the recognition ability of the network for distant targets is improved by supplementing the fusion feature layer and introducing an additional detection head; thirdly, the original detection head is decoupled, and the expression ability of the network for target features is improved by changing the bounding box regression and target classification process into two branches; then, the prior frame is re-clustered, and the height-width ratio of the prior frame is adjusted by the K-means++ algorithm to enhance the network's positioning ability for the target; finally, ablation, comparison and visualization verification experiments are carried out with AP, mAP and FPS as evaluation indicators. The results show that the detection speed of the proposed algorithm on the BDD100K dataset is 95.2 frames/second, and the average accuracy reaches 55.6%, which is higher than that of the YOLOv5 algorithm 6.7%. It can be seen that the improved YOLOv5 algorithm has good target detection accuracy while meeting the real-time requirements of detection. It is suitable for road target detection tasks in complex traffic environments and has guiding significance for improving the visual perception ability of autonomous vehicles.

Automatic location and recognition of horse freezing brand using rotational YOLOv5 deep learning network

Individual livestock identification is of great importance to precision livestock farming. Liquid nitrogen freezing labeled horse brand is an effective way for livestock individual identification. Along with various technological developments, deep-learning-based methods have been applied in such individual marking recognition. In this research, a deep learning method for oriented horse brand location and recognition was proposed. Firstly, Rotational YOLOv5 (R-YOLOv5) was adopted to locate the oriented horse brand, then the cropped images of the brand area were trained by YOLOv5 for number recognition. In the first step, unlike classical detection methods, R-YOLOv5 introduced the orientation into the YOLO framework by integrating Circle Smooth Label (CSL). Besides, Coordinate Attention (CA) was added to raise the attention to positional information in the network. These improvements enhanced the accuracy of detecting oriented brands. In the second step, number recognition was considered as a target detection task because of the requirement of accurate recognition. Finally, the whole brand number was obtained according to the sequences of each detection box position. The experiment results showed that R-YOLOv5 outperformed other rotating target detection algorithms, and the AP (Average Accuracy) was 95.6 %, the FLOPs were 17.4 G, the detection speed was 14.3 fps. As for the results of number recognition, the mAP (mean Average Accuracy) was 95.77 %, the weight size was 13.71 MB, and the detection speed was 68.6 fps. The two-step method can accurately identify brand numbers with complex backgrounds. It also provides a stable and lightweight method for livestock individual identification.

Generation of NIR Spectral Band from RGB Image with Wavelet Domain Spectral Extrapolation Generative Adversarial Network

Near-infrared (NIR) imaging exhibits outstanding penetration capabilities and robust anti-interference characteristics. However, acquiring high-resolution and high-fidelity NIR images is difficult due to the limited mobility, high cost, and low resolution of NIR imaging hardware. In this paper, we proposed a new end-to-end Wavelet Domain Spectral Extrapolation Generative Adversarial Network (WSEGAN) to generate highly realistic NIR images from RGB images. Since RGB and NIR images have different types of noise and artifacts, which affects the quality of NIR images generation. We design a generator with a discrete wavelet transform and an attention mechanism to capture multi-resolution contextual information, and a multi-scale discriminator to capture detailed and global features. Then, the proposed approach is evaluated by three different datasets to achieve optimal results regarding both visual effects and quantitative evaluation. The normalized difference vegetation index (NDVI) is utilized to validate the effectiveness of the generated NIR images. The result demonstrates a strong correlation between the generated images and the actual vegetation distribution. More importantly, the proposed network is testified to generate NIR images to be fused with the RGB image source for agricultural target detection tasks. In dark conditions, results show that using multi-modal data instead of RGB images improves mAP0.5 detection accuracy by 4% on the CAPSICUM dataset and by 8% on the KIWI dataset across five object detection methods. This follows the physical significance of the NIR imaging and demonstrates the potential use of the NIR images generated by the proposed method.

Chunk Duration Limits the Learning of Multiword Chunks: Behavioral and Electroencephalogram Evidence from Statistical Learning.

Language comprehension involves the grouping of words into larger multiword chunks. This is required to recode information into sparser representations to mitigate memory limitations and counteract forgetting. It has been suggested that electrophysiological processing time windows constrain the formation of these units. Specifically, the period of rhythmic neural activity (i.e., slow-frequency neural oscillations) may set an upper limit of 2-3 sec. Here, we assess whether learning of new multiword chunks is also affected by this neural limit. We applied an auditory statistical learning paradigm of an artificial language while manipulating the duration of to-be-learnt chunks. Participants listened to isochronous sequences of disyllabic pseudowords from which they could learn hidden three-word chunks based on transitional probabilities. We presented chunks of 1.95, 2.55, and 3.15 sec that were created by varying the pause interval between pseudowords. In a first behavioral experiment, we tested learning using an implicit target detection task. We found better learning for chunks of 2.55 sec as compared to longer durations in line with an upper limit of the proposed time constraint. In a second experiment, we recorded participants' electroencephalogram during the exposure phase to use frequency tagging as a neural index of statistical learning. Extending the behavioral findings, results show a significant decline in neural tracking for chunks exceeding 3 sec as compared to both shorter durations. Overall, we suggest that language learning is constrained by endogenous time constraints, possibly reflecting electrophysiological processing windows.

A Ghost and Attention Mechanism-Based Deep Learning Approach for SAR Small Target Image Detection

Sy nthetic aperture radar (SAR) images play an important role in the task of small target imaging and detection. However, due to its blurred target imaging, complex image background, and challenging feature representation, SAR detection is difficult to get high precision in a complex imaging environment. Therefore, by integrating the ghost model and the Attention-Mechanism (AM), this paper presents a new feature network model based on the improved YOLOv5s for detecting small targets in SAR images efficiently. The key is to design a Ghost_AM module by integrating the ghost bottleneck with the simple AM. This Ghost_AM module is introduced into to the YOLOv5s’ network for improving the YOLOv5s’ extraction performance for small target’s feature from a complex SAR image. Furthermore, to improve the YOLOv5s’ recognition ability for the features extracted from the Ghost_AM, a bidirectional feature pyramid network is use to replace the YOLOv5s’ feature fusion network and get a precise detection. The experiments using some SAR ship image datasets have demonstrated the presented model is better than some traditional SAR target detection algorithms and the original YOLOv5s. Moreover, the presented model’s parameters are reduced by 25.1% compared with the original YOLOv5s, which contributes to its simpler implementation.

CP-RDM: a new object detection algorithm for casting and pouring robots

Abstract Automating the casting sector heavily relies on pivotal technology for object detection in pouring robots. A sophisticated algorithm designed to identify and locate target pouring holes in intricate casting workshops is crucial for advancing the intelligence of the casting process. However, the workshop environment for pouring is generally challenging, with uneven lighting, varying sizes of pouring holes, and significant occlusion in the target area, all impacting the accuracy of target detection tasks.
To overcome these challenges, this paper proposes enhancing the YOLOv8s algorithm for object detection in pouring robots. Firstly, to address the issue of different scales in pouring holes, a Multi-Scale Residual Channel and Spatial Information Fusion Module (MRCS) is designed to aggregate channel and spatial information, thereby enhancing the feature extraction capability of the model. The proposed enhancement is validated on the Pascal VOC dataset. Secondly, a SimAM attention mechanism is added at the end of the backbone network to focus the object detection network more on the positional region of the pouring hole. Importantly, this addition does not introduce extra parameters or computational burden to the model.
Finally, in the detection part of the model, the detection head from the RT-DETR model is introduced. This combination of real-time detection capability from YOLO and deep feature extraction capability from RT-DETR enhances the detection accuracy of the model while ensuring real-time performance. Experimental results on the updated pouring hole dataset reveal that, with only a slight increase in parameters, the proposed model achieves a 2.5% and 3.5% improvement in mAP@0.5 and F1-Score, respectively, compared to the baseline algorithm YOLOv8s. Precision (P) is enhanced by 1.8%, recall (R) by 3.5%, and PFS reaches 110, meeting the requirements for real-time pouring in pouring robots.

The modulation of temporal predictability on attentional boost effect.

The attentional boost effect, characterized by better memory for background scenes coinciding with a detection target than a nontarget, is believed to stem from a temporary increase in attentional capacity at the time of an acute behavior-related event occurring. Sisk and Jiang's study found that the attentional boost effect also occurs when the target's appearance was predictable. Unfortunately, the duration of the predictive interval in Sisk and Jiang's study was fixed. Since different predictive intervals had different weakening degrees to the acuteness of the target, this fixed duration hindered further investigation into the impact of different levels of predictability on the attentional boost effect. Using the encoding-recognition paradigm and the remembering/knowing paradigm, and setting target stimuli with different predictive interval in target detection tasks, the current study aimed to explore the influence of varying the duration of the predictive interval on the attentional boost effect. The attentional boost effect was observed only in the short and medium predictive duration conditions, but not in the long predictive duration condition. Moreover, as the duration of the predictive interval increased, participants' memory performance on target-paired words gradually declined, while their memory performance on distractor-paired and baseline-paired words gradually improved. Predictability may alter the task demands, allowing participants to more effectively allocate attentional resources to the two tasks at hand.

Pre-stimulus EEG phase coherence predicts visual target detection failures in schizophrenia: A pilot study

Impaired visual target detection is a common finding in schizophrenia that is linked to poor functional outcomes. However, the neural mechanisms that contribute to this deficit remain unclear. Recent research in healthy samples has identified relationships between the phase of pre-stimulus electroencephalographic (EEG) activity in the alpha band (8-12 Hz) or theta band (4-7 Hz) and the likelihood of visual target detection with and without attentional cueing, but these effects have not yet been explored in schizophrenia. We performed a study to investigate such effects in schizophrenia (n = 19) and healthy participants (n = 14), using a visual target detection task with attentional cues. We found significant relationships between pre-stimulus EEG phase properties and visual target detection in both groups, but also clear differences in the effects as a function of frequency, group, and attentional cueing. Alpha-band phase effects were relatively uniform across groups and conditions. By contrast, theta-band phase effects showed differences by group and attentional condition which could be consistent with attentional hyperfocusing in the schizophrenia group. Thus, our results elucidate a novel neural mechanism that may help to explain known impairments affecting both visual target detection and attention in schizophrenia.

MSFE-UIENet: A Multi-Scale Feature Extraction Network for Marine Underwater Image Enhancement

Underwater optical images have outstanding advantages for short-range underwater target detection tasks. However, owing to the limitations of special underwater imaging environments, underwater images often have several problems, such as noise interference, blur texture, low contrast, and color distortion. Marine underwater image enhancement addresses degraded underwater image quality caused by light absorption and scattering. This study introduces MSFE-UIENet, a high-performance network designed to improve image feature extraction, resulting in deep-learning-based underwater image enhancement, addressing the limitations of single convolution and upsampling/downsampling techniques. This network is designed to enhance the image quality in underwater settings by employing an encoder–decoder architecture. In response to the underwhelming enhancement performance caused by the conventional networks’ sole downsampling method, this study introduces a pyramid downsampling module that captures more intricate image features through multi-scale downsampling. Additionally, to augment the feature extraction capabilities of the network, an advanced feature extraction module was proposed to capture detailed information from underwater images. Furthermore, to optimize the network’s gradient flow, forward and backward branches were introduced to accelerate its convergence rate and improve stability. Experimental validation using underwater image datasets indicated that the proposed network effectively enhances underwater image quality, effectively preserving image details and noise suppression across various underwater environments.

Interaction of spatial attention and the associated reward value of audiovisual objects

Reward value and selective attention both enhance the representation of sensory stimuli at the earliest stages of processing. It is still debated whether and how reward-driven and attentional mechanisms interact to influence perception. Here we ask whether the interaction between reward value and selective attention depends on the sensory modality through which the reward information is conveyed. Human participants first learned the reward value of uni-modal visual and auditory stimuli during a conditioning phase. Subsequently, they performed a target detection task on bimodal stimuli containing a previously rewarded stimulus in one, both, or neither of the modalities. Additionally, participants were required to focus their attention on one side and only report targets on the attended side. Our results showed a strong modulation of visual and auditory event-related potentials (ERPs) by spatial attention. We found no main effect of reward value but importantly we found an interaction effect as the strength of attentional modulation of the ERPs was significantly affected by the reward value. When reward effects were examined separately with respect to each modality, auditory value-driven modulation of attention was found to dominate the ERP effects whereas visual reward value on its own led to no effect, likely due to its interference with the target processing. These results inspire a two-stage model where first the salience of a high reward stimulus is enhanced on a local priority map specific to each sensory modality, and at a second stage reward value and top-down attentional mechanisms are integrated across sensory modalities to affect perception.

Adversarial enhancement generation method for side-scan sonar images based on DDPM–YOLO

As marine exploration advances, efficient and accurate seabed target identification becomes increasingly critical. However, traditional methods and current technologies face challenges such as scarce samples and complex imaging conditions when dealing with side-scan sonar images. Given the scarcity of sample augmentation methods for side-scan sonar, this paper iteratively trains the Denoising Diffusion Probabilistic Models(DDPM), integrating the DDPM diffusion model and the downstream You Only Look Once(YOLO) retrieval task into a mutually reinforcing framework, proposing an adversarial enhancement generation method based on the DDPM and YOLO detection models. Experiments demonstrate that the DDPM model generated through this adversarial enhancement generation method can improve the accuracy of downstream YOLO target detection tasks by 7%. The images generated by this model also perform optimally on the Fréchet Inception Distance(FID), Maximum Mean Discrepancy(MMD), and Learned Perceptual Image Patch Similarity(LPIPS) metrics, thereby proving that our method can enhance the quality of generated images from the side-scan sonar diffusion model and offer a new avenue for improving the construction of underwater target detection models.