COCO Dataset Research Articles

Cascade contour-enhanced panoptic segmentation for robotic vision perception

Panoptic segmentation plays a crucial role in enabling robots to comprehend their surroundings, providing fine-grained scene understanding information for robots' intelligent tasks. Although existing methods have made some progress, they are prone to fail in areas with weak textures, small objects, etc. Inspired by biological vision research, we propose a cascaded contour-enhanced panoptic segmentation network called CCPSNet, attempting to enhance the discriminability of instances through structural knowledge. To acquire the scene structure, a cascade contour detection stream is designed, which extracts comprehensive scene contours using channel regulation structural perception module and coarse-to-fine cascade strategy. Furthermore, the contour-guided multi-scale feature enhancement stream is developed to boost the discrimination ability for small objects and weak textures. The stream integrates contour information and multi-scale context features through structural-aware feature modulation module and inverse aggregation technique. Experimental results show that our method improves accuracy on the Cityscapes (61.2 PQ) and COCO (43.5 PQ) datasets while also demonstrating robustness in challenging simulated real-world complex scenarios faced by robots, such as dirty cameras and rainy conditions. The proposed network promises to help the robot perceive the real scene. In future work, an unsupervised training strategy for the network could be explored to reduce the training cost.

Object Detection in Historical Images: Transfer Learning and Pseudo Labelling

The automatic analysis of images in the historical sciences often requires the identification of objects. Object identification is a well researched problem for modern photographs, however, for historical material annotations are often necessary. We present a solution for finding objects without manual work. The method consists of a style transfer of images from the COCO dataset into the domain using CycleGAN and training with items obtained through pseudo labelling on the original and the additional transferred COCO images. Different strategies to assemble the dataset are compared. The best method obtains a F1 score of 0.58 for 15 object types without any labelling.

NAS-BNN: Neural Architecture Search for Binary Neural Networks

Binary Neural Networks (BNNs) have gained extensive attention for their superior inferencing efficiency and compression ratio compared to traditional full-precision networks. However, due to the unique characteristics of BNNs, designing a powerful binary architecture is challenging and often requires significant manpower. A promising solution is to utilize Neural Architecture Search (NAS) to assist in designing BNNs, but current NAS methods for BNNs are relatively straightforward and leave a performance gap between the searched models and manually designed ones. To address this gap, we propose a novel neural architecture search scheme for binary neural networks, named NAS-BNN. We first carefully design a search space based on the unique characteristics of BNNs. Then, we present three training strategies, which significantly enhance the training of supernet and boost the performance of all subnets. Our discovered binary model family outperforms previous BNNs for a wide range of operations (OPs) from 20M to 200M. For instance, we achieve 68.20% top-1 accuracy on ImageNet with only 57M OPs. In addition, we validate the transferability of these searched BNNs on the object detection task, and our binary detectors with the searched BNNs achieve a novel state-of-the-art result, e.g., 31.6% mAP with 370M OPs, on MS COCO dataset. The source code and models will be released at https://github.com/VDIGPKU/NAS-BNN.

ExamGuard: Smart contracts for secure online test

The education sector is currently experiencing profound changes, primarily driven by the widespread adoption of online platforms for conducting examinations. This paper delves into the utilization of smart contracts as a means to revolutionize the monitoring and execution of online examinations, thereby guaranteeing the traceability of evaluation data and examinee activities. In this context, the integration of advanced technologies such as the PoseNet algorithm, derived from the TensorFlow Model, emerges as a pivotal component. By leveraging PoseNet, the system adeptly identifies both single and multiple faces of examinees, thereby ensuring the authenticity and integrity of examination sessions. Moreover, the incorporation of the COCO dataset facilitates the recognition of objects within examination environments, further bolstering the system's capabilities in monitoring examinee activities.of paramount importance is the secure storage of evidence collected during examinations, a task efficiently accomplished through the implementation of the blockchain technology. This platform not only ensures the immutability of data but also safeguards against potential instances of tampering, thereby upholding the credibility of examination results. Through the utilization of smart contracts, the proposed framework not only streamlines the examination process but also instills transparency and integrity, thereby addressing inherent challenges encountered in traditional examination methods. One of the key advantages of this technological integration lies in its ability to modernize examination procedures while concurrently reinforcing trust and accountability within the educational assessment ecosystem. By harnessing the power of smart contracts, educational institutions can mitigate concerns pertaining to data manipulation and malpractice, thereby fostering a more secure and reliable examination environment. Furthermore, the transparency afforded by blockchain technology ensures that examination outcomes are verifiable and auditable, instilling confidence among stakeholders and enhancing the overall credibility of the assessment process. In conclusion, the adoption of smart contracts represents a paradigm shift in the realm of educational assessment, offering a comprehensive solution to the challenges posed by traditional examination methods. By embracing advanced technologies such as PoseNet and blockchain, educational institutions can not only streamline examination procedures but also uphold the highest standards of integrity and accountability. As such, the integration of smart contracts holds immense potential in shaping the future of online examinations, paving the way for a more efficient, transparent, and trustworthy assessment ecosystem.

Nighttime Pothole Detection: A Benchmark

In the field of computer vision, the detection of road potholes at night represents a critical challenge in enhancing the safety of intelligent transportation systems. Ensuring road safety is of paramount importance, particularly in promptly repairing pothole issues. These abrupt road depressions can easily lead to vehicle skidding, loss of control, and even traffic accidents, especially when water has pooled in or submerged the potholes. Therefore, the detection and recognition of road potholes can significantly reduce vehicle damage and the incidence of safety incidents. However, research on road pothole detection lacks high-quality annotated datasets, particularly under low-light conditions at night. To address this issue, this study introduces a novel Nighttime Pothole Dataset (NPD), independently collected and comprising 3831 images that capture diverse scene variations. The construction of this dataset aims to counteract the insufficiency of existing data resources and strives to provide a richer and more realistic benchmark. Additionally, we develop a baseline detector, termed WT-YOLOv8, for the proposed dataset, based on YOLOv8. We also evaluate the performance of the improved WT-YOLOv8 method and eight state-of-the-art object detection methods on the NPD and the COCO dataset. The experimental results on the NPD demonstrate that WT-YOLOv8 achieves a 2.3% improvement in mean Average Precision (mAP) over YOLOv8. In terms of the key metrics—AP@0.5 and AP@0.75—it shows enhancements of 1.5% and 2.8%, respectively, compared to YOLOv8. The experimental results provide valuable insights into each method’s strengths and weaknesses under low-light conditions. This analysis highlights the importance of a specialized dataset for nighttime pothole detection and shows variations in accuracy and robustness among methods, emphasizing the need for improved nighttime pothole detection techniques. The introduction of the NPD is expected to stimulate further research, encouraging the development of advanced algorithms for nighttime pothole detection, ultimately leading to more flexible and reliable road maintenance and road safety.

LVD-YOLO: An efficient lightweight vehicle detection model for intelligent transportation systems

Vehicle detection is a fundamental component of intelligent transportation systems. However, current algorithms often encounter issues such as high computational complexity, long execution times, and significant resource demands, making them unsuitable for resource-limited environments. To overcome these challenges, we propose LVD-YOLO, a Lightweight Vehicle Detection Model based on YOLO. This model incorporates the EfficientNetv2 network structure as its backbone, which reduces parameters and enhances feature extraction capabilities. By utilizing a bidirectional feature pyramid structure along with a dual attention mechanism, we enable efficient information exchange across feature layers, thereby improving multiscale feature fusion. Additionally, we refine the model's loss function with SIoU loss to boost regression and prediction performance. Experimental results on the PASCAL VOC and MS COCO datasets show that LVD-YOLO outperforms YOLOv5s, achieving a 0.5% increase in accuracy while reducing FLOPs by 64.6% and parameters by 48.6%. These improvements highlight its effectiveness for use in resource-constrained environments.

Innovation in public health surveillance for social distancing during the COVID-19 pandemic: A deep learning and object detection based novel approach.

The Corona Virus Disease (COVID-19) has a huge impact on all of humanity, and people's disregard for COVID-19 regulations has sped up the disease's spread. Our study uses a state-of-the-art object detection model like YOLOv4 (You Only Look Once, version 4), a very effective tool, on real-time 25fps, 1920 X 1080 video data streamed live by a camera-mounted Unmanned Aerial Vehicle (UAV) quad-copter to observe proper maintenance of social distance in an area of 35m range in this study. The model has demonstrated remarkable efficacy in identifying and quantifying instances of social distancing, with an accuracy of 82% and little latency. It has been able to work efficiently with real-time streaming at 25-30 ms. Our model is based on CSPDarkNet-53, which was trained on the MS COCO dataset for image classification. It includes additional layers to capture feature maps from different phases. Additionally, the model's neck is made up of PANet, which is used to aggregate the parameters from various CSPDarkNet-53 layers. The CSPDarkNet-53's 53 convolutional layers are followed by 53 more layers in the model head, for a total of 106 completely convolutional layers in the design. This architecture is further integrated with YOLOv3, resulting in the YOLOv4 model, which will be used by our detection model. Furthermore, to differentiate humans The aforementioned method was used to evaluate drone footage and count social distance violations in real time. Our findings show that our model was reliable and successful at detecting social distance violations in real-time with an average accuracy of 82%.

Novel concept-based image captioning models using LSTM and multi-encoder transformer architecture

Captioning an image involves using a combination of vision and language models to describe the image in an expressive and concise sentence. Successful captioning task requires extracting as much information as possible from the corresponding image. One of these key pieces of information is the topic to which the image belongs. The state-of-the-art methods used topic modeling depending only on caption text in order to extract these topics. The problem with extracting the topics using topic modeling only on caption text is that it lacks the consideration of the image’s semantic information. Instead, concept modeling extracts the concepts directly from the images in addition to considering the corresponding caption text. Concept modeling can be used in image captioning to extremely capture the image contexts and benefit from it to produce more accurate descriptions. In this paper, novel image captioning models are proposed by utilizing the concept modeling technique. The first concept-based model is proposed by utilizing LSTM as a decoder while the second model is proposed in association with new multi-encoder transformer architecture. Standard metrics have been used to evaluate the proposed models using Microsoft COCO and Flickr30K datasets. The proposed models outperformed the related work methods with reduced computational complexity.

Optimizing Automated Optical Inspection: An Adaptive Fusion and Semi-Supervised Self-Learning Approach for Elevated Accuracy and Efficiency in Scenarios with Scarce Labeled Data.

In the field of automatic optical inspection (AOI), this study presents innovative strategies to enhance object detection accuracy while minimizing dependence on large annotated datasets. We initially developed a defect detection model using a dataset of 3579 images across 32 categories, created in collaboration with a major Taiwanese panel manufacturer. This model was evaluated using 12,000 ambiguously labeled images, with improvements achieved through data augmentation and annotation refinement. To address the challenges of limited labeled data, we proposed the Adaptive Fused Semi-Supervised Self-Learning (AFSL) method. This approach, designed for anchor-based object detection models, leverages a small set of labeled data alongside a larger pool of unlabeled data to enable continuous model optimization. Key components of AFSL include the Bounding Box Assigner, Adaptive Training Scheduler, and Data Allocator, which together facilitate dynamic threshold adjustments and balanced training, significantly enhancing the model's performance on AOI datasets. The AFSL method improved the mean average precision (mAP) from 43.5% to 57.1% on the COCO dataset and by 2.6% on the AOI dataset, demonstrating its effectiveness in achieving high levels of precision and efficiency in AOI with minimal labeled data.

Learning General and Specific Embedding with Transformer for Few-Shot Object Detection

AbstractFew-shot object detection (FSOD) studies how to detect novel objects with few annotated examples effectively. Recently, it has been demonstrated that decent feature embeddings, including the general feature embeddings that are more invariant to visual changes and the specific feature embeddings that are more discriminative for different object classes, are both important for FSOD. However, current methods lack appropriate mechanisms to sensibly cooperate both types of feature embeddings based on their importance to detecting objects of novel classes, which may result in sub-optimal performance. In this paper, to achieve more effective FSOD, we attempt to explicitly encode both general and specific feature embeddings using learnable tensors and apply a Transformer to help better incorporate them in FSOD according to their relations to the input object features. We thus propose a Transformer-based general and specific embedding learning (T-GSEL) method for FSOD. In T-GSEL, learnable tensors are employed in a three-stage pipeline, encoding feature embeddings in general level, intermediate level, and specific level, respectively. In each stage, we apply a Transformer to first model the relations of the corresponding embedding to input object features and then apply the estimated relations to refine the input features. Meanwhile, we further introduce cross-stage connections between embeddings of different stages to make them complement and cooperate with each other, delivering general, intermediate, and specific feature embeddings stage by stage and utilizing them together for feature refinement in FSOD. In practice, a T-GSEL module is easy to inject. Extensive empirical results further show that our proposed T-GSEL method achieves compelling FSOD performance on both PASCAL VOC and MS COCO datasets compared with other state-of-the-art approaches.

Wearable damaged clothing fabric image recognition system based on image restoration algorithm

With the improvement of people's living standards and the rapid growth of national technological strength, wearable devices have gradually integrated into people's daily lives, making them convenient for people's work and lives. Traditional fabric detection methods often rely on manual visual inspection, which has low efficiency and accuracy that does not meet the requirements. In view of this, the research has innovatively integrated GAN, CNN, field Programmable gate array (FPGA), multi-stage image recovery algorithm based on graph neural network, for high frequency and low frequency fusion, and combined it with wearable devices to develop a new wearable fabric recovery and recognition system. The results showed that in the COCO dataset, after 250 iterations, the WFR2S algorithm achieved a damage identification accuracy of 95.5 %, a mean absolute error of 0.011, a real-time processing speed of 30 fps, and a resource consumption of 2.4 W. These data demonstrate the significant performance advantages of WFR2S in image recovery and damage detection. The newly developed image recovery algorithm improves the accuracy, availability and stability of damage recognition in garment fabrics. This study not only has theoretical value but also has broad application prospects in practical application. The aim is to contribute to the intelligent development of the garment industry.

Object detection with a dynamic interactive network based on relational graph routing

Combinatorial relational reasoning in neural networks used for object detection is usually static; therefore, it cannot selectively fuse visual information and semantic relations, which limits their performance. To address this problem, we propose a relational graph routing network (RGRN) that enables the dynamic interaction of visual and semantic features. The network consists of a dynamic graph network, dual path-sharing module, and relational routing interaction module. First, we used a data-driven technique to obtain the semantic information between tags from the dataset. Rich semantic information was obtained by calculating the similarity between tags. Second, the two types of semantic information were fused using a dynamic graph network to capture high-level semantic information. The visual and semantic features are then filtered and encoded through the dual path-sharing module to obtain enhanced visual and semantic features. Finally, three units were used to dynamically fuse visual and semantic information in the relational routing interaction module, which densely links the three units and routers to construct a routing space that can autonomously decide on the optimal fusion path through model learning. A series of experiments was conducted on the MS COCO dataset. RGRN achieved 54.7% box AP on object detection, which was 2.8% box AP higher than that of the Cascade Mask R-CNN. The experimental results show that the routing space enables better interaction between visual and semantic information. Therefore, our method can achieve better performance than many state-of-the-art methods.

Content-based image retrieval by classification with reinforcement optimisation evolutionary machine learning with applications

ABSTRACT Content Based Image Retrieval (CBIR) plays a significant role in identifying the similarity of images with large datasets. It is identified based on the size, colour, and texture features of the image. But in such conditions, it is complex to determine the features of query images in large datasets and does not show accurate similarity when compared with every image in the retrieval process. In order to perform an efficient similarity of images, a novel Machine Learning (ML) approach Kernelized Radial Basis Auto-Encoder Function Neural Network (Ker_RadBAEFNN) technique is proposed that performs the individual image classification in the retrieval process. Moreover, the neural networks are optimised based on the reinforcement process and perform the extraction process regarding individual images. Further,reinforcement-based optimisation estimates the images in neural networks for undertaking an automatic feature extraction of query images. The performance of the classification process is validated based on MNIST, METU, and COCO datasets that determined the efficiency of the recognition and classification process of image retrieval. The experimental analysis is carried out based on various measures such as accuracy, precision, recall, F1-score, RMSE, and MAPE for the proposed and existing GLCM-ABC, PSO-ANN, IRB-CNN, FAGWO, and OCAM methods. The analysis shows that the performance of the proposed attained better effectiveness with attained accuracy by 98% and diminished for state-of-the-art techniques as 92%, 95%, 94%, 96.8%, as well as 96%, respectively. Compared to existing methods, the accuracy rate of the proposed method is maximised by 1.3%.

Multi-Dimensional Fusion Attention Mechanism with Vim-like Structure for Mobile Network Design

Recent advancements in mobile neural networks, such as the squeeze-and-excitation (SE) attention mechanism, have significantly improved model performance. However, they often overlook the crucial interaction between location information and channels. The interaction of multiple dimensions in feature engineering is of paramount importance for achieving high-quality results. The Transformer model and its successors, such as Mamba and Vision Mamba, have effectively combined features and linked location information. This approach has transitioned from NLP (natural language processing) to CV (computer vision). This paper introduces a novel attention mechanism for mobile neural networks inspired by the structure of Vim (Vision Mamba). It adopts a “1 + 3” architecture to embed multi-dimensional information into channel attention, termed ”Multi-Dimensional Vim-like Attention Mechanism”. The proposed method splits the input into two major branches: the left branch retains the original information for subsequent feature screening, while the right branch divides the channel attention into three one-dimensional feature encoding processes. These processes aggregate features along one channel direction and two spatial directions, simultaneously capturing remote dependencies and preserving precise location information. The resulting feature maps are then combined with the left branch to produce direction-aware, location-sensitive, and channel-aware attention maps. The multi-dimensional Vim-like attention module is simple and can be seamlessly integrated into classical mobile neural networks such as MobileNetV2 and ShuffleNetV2 with minimal computational overhead. Experimental results demonstrate that this attention module adapts well to mobile neural networks with a low parameter count, delivering excellent performance on the CIFAR-100 and MS COCO datasets.

Incorporating long-tail data in complex backgrounds for visual surface defect detection in PCBs

AbstractHigh-quality printed circuit boards (PCBs) are essential components in modern electronic circuits. Nevertheless, most of the existing methods for PCB surface defect detection neglect the fact that PCB surface defects in complex backgrounds are prone to long-tailed data distributions, which in turn affects the effectiveness of defect detection. Additionally, most of the existing methods ignore the intra-scale features of defects and do not utilize auxiliary supervision strategies to improve the detection performance of the network. To tackle these issues, we propose a lightweight long-tailed data mining network (LLM-Net) for identifying PCB surface defects. Firstly, the proposed Efficient Feature Fusion Network (EFFNet) is applied to embed intra-scale feature associations and multi-scale features of defects into LLM-Net. Next, an auxiliary supervision method with a soft label assignment strategy is designed to help LLM-Net learn more accurate defect features. Finally, the issue of inadequate tail data detection is addressed by employing the devised Binary Cross-Entropy Loss Rank Mining method (BCE-LRM) to identify challenging samples. The performance of LLM-Net was evaluated on a homemade dataset of PCB surface soldering defects, and the results show that LLM-Net achieves the best accuracy of mAP@0.5 for the evaluation metric of the COCO dataset, and it has a real-time inference speed of 188 frames per second (FPS).

Two-stage re-parameterization and sample disentanglement network for surveillance vehicle detection

Abstract Detecting vehicles from a surveillance viewpoint is essential, as it has wide applications in community security and traffic control. However, existing methods completely overlook the high memory access costs (MAC) and low degree of parallelism inherent in multi-branch topologies, resulting in significant latency during inference. Additionally, existing methods share the same positive/negative sample set between the classification and localization branches, leading to sample misalignment, and rely solely on intersection-over-union (IoU) for sample assignment, thereby causing a decrease in detection performance. To tackle these issues, this paper introduces a two-stage re-parameterization and sample disentanglement network (TRSD-Net). It is based on two-stage depthwise to pointwise re-parameterization (RepTDP) and task-aligned sample disentanglement (TSD). RepTDP employs structural re-parameterization to decouple the multi-branch topology during training and the plain architecture during inference, thus achieving low latency. By employing different sample assignment strategies, TSD can adaptively select the most suitable positive/negative sample sets for classification and localization tasks, thereby enhancing detection performance. Additionally, TSD considers three important factors influencing sample assignment. TRSD-Net is evaluated on both the UA-DETRAC and COCO datasets. On the UA-DETRAC dataset, compared to state-of-the-art (SOTA) methods, TRSD-Net improves the detection accuracy from 58.8% to 59.7%. Additionally, it reduces the parameter count by 87%, the computational complexity by 85%, and the latency by 39%, while increasing the detection speed by 65%. Similar performance improvement trends are observed on the COCO dataset.

Improved research on coral bleaching detection model based on FCOS model

Coral bleaching detection is vital for assessing coral reef health. This paper introduces FCOS_EfficientNET, an improved model that enhances accuracy, recall, and real-time performance in coral bleaching detection. Utilizing EfficientNet as the backbone, the model optimizes parameter throughput. We adopt the ReLU activation function and utilize cosine similarity and softmax to assign weights to datasets, modifying the attention structure to reduce memory consumption. The model also integrates BiFPN for better feature extraction and employs an improved training method to enhance detection accuracy. To cater to different scenarios, we have developed four variants: FCOS_EfficientNETb0, FCOS_EfficientNETb1, FCOS_EfficientNETb2, and FCOS_EfficientNETb3. Experimental results on the MS COCO dataset show that FCOS_EfficientNETb3 achieves a mean average precision (mAP) of 48.5%, while FCOS_EfficientNETb0 reaches a frame rate of 167.17 fps, highlighting the superior performance of the series. On a custom coral bleaching detection dataset, FCOS_EfficientNETb3 achieves 81.5% accuracy and a 59.3% recall rate, demonstrating the effectiveness of the model. FCOS_EfficientNETb1 and FCOS_EfficientNETb2 offer a balance between operations per second, frame rate, and mAP, making them suitable for mobile and edge computing. These models effectively track movement or changes in marine traffic around coral reefs with moderate fps and recall rates.

Pulmonary nodule detection in low dose computed tomography using a medical-to-medical transfer learning approach.

Lung cancer is the second most common cancer and the leading cause of cancer death globally. Low dose computed tomography (LDCT) is the recommended imaging screening tool for the early detection of lung cancer. A fully automated computer-aided detection method for LDCT will greatly improve the existing clinical workflow. Most of the existing methods for lung detection are designed for high-dose CTs (HDCTs), and those methods cannot be directly applied to LDCTs due to domain shifts and inferior quality of LDCT images. In this work, we describe a semi-automated transfer learning-based approach for the early detection of lung nodules using LDCTs. In this work, we developed an algorithm based on the object detection model, you only look once (YOLO) to detect lung nodules. The YOLO model was first trained on CTs, and the pre-trained weights were used as initial weights during the retraining of the model on LDCTs using a medical-to-medical transfer learning approach. The dataset for this study was from a screening trial consisting of LDCTs acquired from 50 biopsy-confirmed lung cancer patients obtained over 3 consecutive years (T1, T2, and T3). About 60 lung cancer patients' HDCTs were obtained from a public dataset. The developed model was evaluated using a hold-out test set comprising 15 patient cases (93 slices with cancerous nodules) using precision, specificity, recall, and F1-score. The evaluation metrics were reported patient-wise on a per-year basis and averaged for 3 years. For comparative analysis, the proposed detection model was trained using pre-trained weights from the COCO dataset as the initial weights. A paired t-test and chi-squared test with an alpha value of 0.05 were used for statistical significance testing. The results were reported by comparing the proposed model developed using HDCT pre-trained weights with COCO pre-trained weights. The former approach versus the latter approach obtained a precision of 0.982 versus 0.93 in detecting cancerous nodules, specificity of 0.923 versus 0.849 in identifying slices with no cancerous nodules, recall of 0.87 versus 0.886, and F1-score of 0.924 versus 0.903. As the nodule progressed, the former approach achieved a precision of 1, specificity of 0.92, and sensitivity of 0.930. The statistical analysis performed in the comparative study resulted in a -value of 0.0054 for precision and a -value of 0.00034 for specificity. In this study, a semi-automated method was developed to detect lung nodules in LDCTs using HDCT pre-trained weights as the initial weights and retraining the model. Further, the results were compared by replacing HDCT pre-trained weights in the above approach with COCO pre-trained weights. The proposed method may identify early lung nodules during the screening program, reduce overdiagnosis and follow-ups due to misdiagnosis in LDCTs, start treatment options in the affected patients, and lower the mortality rate.

BSMH: Cross‐dataset object detection based on box‐separated multiple‐head

AbstractCross‐dataset object detection methods can adapt to the needs of rapid category expansion in object detection tasks. However, these methods are prone to generate dataset‐aware errors with false alarm objects. This study is aimed to address these issues. A box‐separated multiple‐head module and box‐separated loss function based on the YOLOv8 network are devised to achieve cross‐dataset object detection. Additionally, a sameclass‐aware fusion module to avoid gradient conflicts due to cross‐category conflicts is developed. A multiple‐head fusion module is devised to reduce the number of false alarm objects caused by dataset‐aware errors. A global class‐aware sampler is also designed to adapt to the impact of the imbalanced number of categories and training samples across datasets. The effectiveness of the box‐separated multiple‐head module is verified using cross‐datasets built using the COCO, WiderFace, WiderPerson, and OpenImages V4 datasets. Extensive experiments demonstrate the efficiency and precision of the proposed method.

IML-SSOD: Interconnected and multi-layer threshold learning for semi-supervised detection

While semi-supervised anchored detector of the R-CNN series has achieved remarkable success, semi-supervised anchor-free detector lacks the ability to generate high-quality flexible pseudo labels, resulting in serious inconsistencies in SSOD. In order to make the network learn more reliable and consistent label data to solve the problem of information bias, we propose an interconnected and multi-layer threshold learning for semi-supervised object detection (IML-SSOD). The Joint Guided Estimation (JGE) module uses the Core Zone refinement module to improve the position accuracy score of low semantic information, and combines the classification and the centerness score as evaluation criteria to predict stable labels. The multi-layer threshold filtering method selects more potential label samples for the student network ensuring the information used in training. Extensive experiments on MS COCO and PASCAL VOC datasets demonstrated the effectiveness of IML-SSOD. Compared with existing methods, our method on VOC achieved 81.9% AP50 and 57.89% AP50:95, which is highly competitive.