Strict NMS: Pedestrian Detection in Crowd Scenes

The non-maximum suppression (NMS) algorithm, which merges neighboring bounding boxes as the detection result, is widely utilized in object detection methods. However, the traditional NMS algorithm is not suitable for text detection in natural scenes, especially for long texts and dense texts. In this paper, we observe that the coordinates of candidate bounding boxes are presented in skew distribution. On this observation, an improved NMS algorithm called SD-NMS (Skew Distribution NMS) is designed. First, the mode and the median of the coordinate set are counted to filter out the redundant bounding boxes. Then the left bounding boxes are merged for the location of the text regions. The SD-NMS method improves the detection ability of the model without extra model training and can be easily embedded in text detection methods. The experimental results show that our method obtains F-measure more than other NMS methods in public data sets ICDAR2015 and MSRATD500.

Non-maximum suppression (NMS) is an indispensable post-processing step in object detection. After a picture is detected by the target, a large number of redundant anchors will be obtained, so we need to go through NMS to process these repeated anchors. Its essence is to extract the target detection frame with high confidence and suppress the false detection frame with low confidence. The biggest problem in the NMS algorithm is that it completely removes adjacent low-confidence detection frames. In some images with dense targets, this is very likely to cause missed detection and false detection. Therefore, we proposed the Simple-NMS algorithm, which adds two new thresholds to the original NMS, and sets a constraint condition different from the original based on the new threshold, and does not change the complexity of the traditional NMS algorithm, which can be very good Improve the effectiveness of NMS missed targets. The new NMS algorithm is improved on the standard data sets PASAL VOC2007 and MS COCO. In addition, the algorithm is simple and efficient, and can be easily integrated into any other object detection process.

While remarkable progress has been made to pedestrian detection in recent years, robust pedestrian detection in the wild e.g., under surveillance scenarios with occlusions, remains a challenging problem. In this paper, we present a novel approach for joint pedestrian and body part detection via semantic relationship learning under unconstrained scenarios. Specifically, we propose a Body Part Indexed Feature (BPIF) representation to encode the semantic relationship between individual body parts (i.e., head, head-shoulder, upper body, and whole body) and highlight per body part features, providing robustness against partial occlusions to the whole body. We also propose an Adaptive Joint Non-Maximum Suppression (AJ-NMS) to replace the original NMS algorithm widely used in object detection, leading to higher precision and recall for detecting overlapped pedestrians. Experimental results on the public-domain CUHK-SYSU Person Search Dataset show that the proposed approach outperforms the state-of-the-art methods for joint pedestrian and body part detection in the wild.

Anomaly detection technology in crowd scene is very important in public place. Crowd detection differs from pedestrian detection which we assume no individual pedestrian can be properly segmented in the image. We propose a scheme which the scen can be treated the crowd motion patterns as the spatial-temporal domain. In the classification stage, we divide whole frame into small blocks, and motion pattern in each block is encoded by the distribution of motion bags in it. PLSA classifier is proposed to infer classification of crowed detection, and we classify motion pattern into normal or abnormal group according to the deviation between motion pattern and train model. The comprehensive implementation can detect crowd in real-time. This paper presents an approach to automatically detect abnormal behavior in crowd scene with Interest points to represent moving objects to generate word of bags, which are used to describe crowed moriment results show that the speed of detection has been greatly improved using our approach.

Pattern recognition of defects in wafer drawing is very important to determine the root cause of production defects, which can provide insights for the production improvement of wafer plants. In wafer manufacturing process, there will be mixed defects, detection of mixed defects is much more complex. Deep learning method has become a good choice to solve the problem of complex pattern recognition. In this paper, an improved Mask R-CNN model for wafer defect map is proposed. The ability of small target recognition is improved by fusing residual network and feature pyramid network, and the target feature can be fully extracted in the shallow model, so that the efficiency of feature extraction is improved. Secondly, an improved non-maximum suppression (NMS) algorithm is added to solve the error detection problem caused by traditional algorithms. The experimental results show that the improved Mask R-CNN model is superior to the conventional model, and has better recognition effect for small targets and shallow models. Secondly, the improved NMS algorithm is added to improve the detection precision.

Despite significant recent progress in pedestrian detection, detecting individuals in crowded scenes remains a challenging task due to the limitations of global detection methods, which heavily rely on Non-Maximum Suppression (NMS) and often result in errors in dense environments. In this paper, we propose a novel two-stage framework to enhance robustness in such scenarios. The first stage employs Cascade R-CNN to detect isolated pedestrians and generate proposal regions in crowded areas, leveraging overlaps for efficient localization. The second stage utilizes a bottom-up pose estimation network, DEKR, to extract human keypoints from these regions, followed by the construction of pseudo-graph images based on topological connections. These pseudo-graphs are then regressed into pedestrian bounding boxes using the YOLOX model, effectively decoupling individuals even with inaccurate keypoints. This approach mitigates NMS shortcomings, achieving a balance between effectiveness and computational efficiency. We evaluate our method on high-density datasets, including Citypersons and a self-constructed Duke-v1 dataset, with extensive experimental results confirming its superior performance and general applicability.

Detecting pedestrians in a crowded scene in real time is a challenging task in monitoring and managing crowd. Many researchers around the world have addressed this task and managed to achieve satisfactory results. However, the problem of automating detection of pedestrians in the crowd is still an open issue depending on the density of crowd in a scene. To ensure safety and security, automating the crowd detection and tracking process in real time is necessary in designing a robust and secure system. Detecting and localizing objects has successfully aided in identifying the major problems with detecting pedestrians and has been a major step forward in managing crowd automatically. In this paper, we have used tiny YOLOv4. YOLO (You Only Look Once) has proved quite useful in detecting and localizing objects in an image with impressive response speed. YOLO network usually scales an entire image into fixed sized grids and then identifies and detects the region into these grids using bounding boxes. Using transfer learning on an already trained YOLO inception model on COCO dataset, detection of pedestrians in surveillance videos is handled. The paper discusses the implementation and detection performance of the proposed YOLOv4 tiny model on the UCSD pedestrian Detection dataset with promising results.

Pedestrian detection is a challenging and hot research topic in the field of computer vision, especially for the crowded scenes where occlusion happens frequently. In this paper, we propose a novel AutoPedestrian scheme that automatically augments the pedestrian data and searches for suitable loss functions, aiming for better performance of pedestrian detection especially in crowded scenes. To our best knowledge, it is the first work to automatically search the optimal policy of data augmentation and loss function jointly for the pedestrian detection. To achieve the goal of searching the optimal augmentation scheme and loss function jointly, we first formulate the data augmentation policy and loss function as probability distributions based on different hyper-parameters. Then, we apply a double-loop scheme with importance-sampling to solve the optimization problem of data augmentation and loss function types efficiently. Comprehensive experiments on two popular benchmarks of CrowdHuman and CityPersons show the effectiveness of our proposed method. In particular, we achieve 40.58% in MR on CrowdHuman datasets and 11.3% in MR on CityPersons reasonable subset, yielding new state-of-the-art results on these two datasets.

Pedestrian Detection (PD) is one of the most challenging problems for advanced driving assistance systems. Although deep models have achieved significant performance, their limited generalization ability prevents real-life implementation. The reason behind this limitation is shown to be the data-dependency of the deep learning, especially when model architecture and parameters are tweaked to boost the performance. As a result, pre-trained deep models tend to perform much worse when they are tested on a different dataset. By focusing on this issue, this paper investigates the use of ensembles consisting of vanilla-style deep models as weak classifiers. The main aim is to investigate if multiple models are used by their default parameters (without any fine-tuning considering the test set performance); would their ensemble reach better generalization capacity and have consistent performance on different datasets? Designing an ensemble for PD significantly differs from conventional classifier ensembles due to the importance of bounding box properties (location, size etc.) found by the detector. Thus, an extensive experimentation is carried out on how to combine model score with bounding box properties. Both simple ensemble strategies and learning based methods have been tested to present a simple and effective pipeline. Finally, non-maximum suppression, which is currently being used for bounding box regression, is modified to develop an ensemble method and the results show that it provides better improvements compared to existing ensemble strategies.

The problem of re-identification of people in a crowd commonly arises in real application scenarios, yet it has received less attention than it deserves. To facilitate research focusing on this problem, we have embarked on constructing a new person re-identification dataset with many instances of crowded indoor and outdoor scenes. This paper proposes a two-stage robust method for pedestrian detection in a complex crowded background to provide bounding box annotations. The first stage is to generate pedestrian proposals using Faster R-CNN and locate each pedestrian using Non-maximum Suppression (NMS). Candidates in dense proposal regions are merged to identify crowd patches. We then apply a bottom-up human pose estimation method to detect individual pedestrians in the crowd patches. The locations of all subjects are achieved based on the bounding boxes from the two stages. The identity of the detected subjects throughout each video is then automatically annotated using multiple features and spatial-temporal clues. The experimental results on a crowded pedestrians dataset demonstrate the effectiveness and efficiency of the proposed method.

Pedestrian detection in crowded scenes is widely used in computer vision. However, it still has two difficulties: 1) eliminating repeated predictions (multiple predictions corresponding to the same object); 2) false detection and missing detection due to the high scene occlusion rate and the small visible area of detected pedestrians. This paper presents a detection framework based on DETR (detection transformer) to address the above problems, and the model is called AD-DETR (asymmetrical relation detection transformer). We find that the symmetry in a DETR framework causes synchronous prediction updates and duplicate predictions. Therefore, we propose an asymmetric relationship fusion mechanism and let each query asymmetrically fuse the relative relationships of surrounding predictions to learn to eliminate duplicate predictions. Then, we propose a decoupled cross-attention head that allows the model to learn to restrict the range of attention to focus more on visible regions and regions that contribute more to confidence. The method can reduce the noise information introduced by the occluded objects to reduce the false detection rate. Meanwhile, in our proposed asymmetric relations module, we establish a way to encode the relative relation between sets of attention points and improve the baseline. Without additional annotations, combined with the deformable-DETR with Res50 as the backbone, our method can achieve an average precision of 92.6%, MR$ ^{-2} $ of 40.0% and Jaccard index of 84.4% on the challenging CrowdHuman dataset. Our method exceeds previous methods, such as Iter-E2EDet (progressive end-to-end object detection), MIP (one proposal, multiple predictions), etc. Experiments show that our method can significantly improve the performance of the query-based model for crowded scenes, and it is highly robust for the crowded scene.

In view of the time-consuming and laborious manual picking and sorting of strawberries, the direct impact of image recognition accuracy on automatic picking and the rapid development of deep learning(DL), a Faster Regions with Convolutional Neural Network features (R-CNN) strawberry recognition method that combines Mixup data augmentation, a ResNet(Residual Network)50 backbone feature extraction network and a Soft-NMS (Non-Maximum Suppression) algorithm, named the MRS Faster R-CNN, is proposed. In this paper, the transfer learning backbone feature extraction network VGG (Visual Geometry Group) 16 and ResNet50 are compared, and the superior ResNet50 is selected as the backbone network of MRS Faster R-CNN. The data augmentation method of Mixup image fusion is used to improve the learning and generalization ability of the model. The redundant bboxes (bounding boxes) are removed through Soft-NMS to obtain the best region proposal. The freezing phase is added to the training process, effectively reducing the occupation of video memory and shortening the training time. After experimental verification, the optimized model improved the AP (Average Precision) values of mature and immature strawberries by 0.26% and 5.34%, respectively, and the P(Precision) values by 0.81% and 6.34%, respectively, compared to the original model (R Faster R-CNN). Therefore, the MRS Faster R-CNN model proposed in this paper has great potential in the field of strawberry recognition and maturity classification and improves the recognition rate of small fruit and overlapping occluded fruit, thus providing an excellent solution for mechanized picking and sorting.

Relay protection devices are necessary to guard the power system safety and stability. With the significant number of substations and relay protection devices, the maintenance workload has become difficult due to limited manpower, posing hidden dangers to the reliable operation of protection devices. In view of this, this article proposes an enhanced Faster R-CNN algorithm to diagnose the relay protection devices based on image monitoring. The proposed algorithm uses RestNet50 as the main tool to recognize features from input images to generate feature maps. Meanwhile, to improve the image detection accuracy, the Non-Maximum Sup-pression (NMS) algorithm in regional suggestion network (RPN) is optimized from solid thresh-old to soft threshold. By combining the images captured by the surveillance camera, intelligent inspections are conducted on the status of the pressing plate, fiber bending, external object retention, and other items in the relay protection device. A great number of samples are used to train networks for different tasks. The results indicate that the enhanced Faster R-CNN has a higher image recognition rate, and the recognition accuracy of each inspection item for the protection device is higher than 95%.

Text in scene images usually contains significant information. Text detection and recognition in the scene is important for a variety of advanced machine vision applications, such as image and video retrieval, automotive assistance, and multilingual translation. In particular, most text recognition systems require texts to be localized in images beforehand and this is a significant demand. The purpose of this study is to provide a method to detect texts in natural images. The proposed approach combines advantages of extremal region, ER, methods and classification of convolutional neural network, CNN. This significantly reduces the false positives and increases the accuracy of detection. The method of sliding windows is employed with different sizes in order to determine text candidates. Extraction of enhanced ERs is performed in three consecutive stages on three distinct color channels, R, G, and B. Then, the results are combined together by an add method. After grouping, the word candidates are classified to two classes of text and non-text sections by a CNN classifier. By applying non-maximum suppression (NMS) algorithm to the same words, words with the highest probability are selected. The average values of accuracy, recall, precision and F-measure of the proposed text detection model on the ICDAR2013 database are 0.893, 0.962, 0.948, and 0.955, respectively. The optimal cut point of the proposed method is 0.648, which has the highest average accuracy, 91.93%. The AUC of ROC and PR diagrams for the proposed model are 0.851 and 0.718, respectively. These results of AUC for ROC and PR curves showed an outstanding enhancement in comparison with the best detection rate of previous methods. Experimental results on the ICDAR2011, ICDAR2013 and ICDAR2015 databases also demonstrate that our algorithm outperforms the state-of-the-art scene text detection methods.

This study addresses the challenge of efficient fruit detection and accurate classification in complex environments. Based on the YOLOv5 framework, a training set was constructed by combining publicly available and self-collected datasets, focusing on three types of fruits: bananas, apples, and oranges. The study incorporated data augmentation, transfer learning, the CBAM (Convolutional Block Attention Module) attention mechanism, and an improved Non-Maximum Suppression (NMS) algorithm. Experimental results demonstrated an average precision of 96.5% and an inference speed of 39.26 frames per second under complex scenarios. Coupled with a user-friendly interface design, this system provides an effective real-time detection and statistical solution for agriculture, retail, and food processing industries.