RPN Network Research Articles

A Few-Shot Object Detection Method for Endangered Species

Endangered species detection plays an important role in biodiversity conservation and is significant in maintaining ecological balance. Existing deep learning-based object detection methods are overly dependent on a large number of supervised samples, and building such endangered species datasets is usually costly. Aiming at the problems faced by endangered species detection, such as low accuracy and easy loss of location information, an efficient endangered species detection method with fewer samples is proposed to extend the few-shot object detection technique to the field of endangered species detection, which requires only a small number of training samples to obtain excellent detection results. First, SE-Res2Net is proposed to optimize the feature extraction capability. Secondly, an RPN network with multiple attention mechanism is proposed. Finally, for the classification confusion problem, a weighted prototype-based comparison branch is introduced to construct weighted category prototype vectors, which effectively improves the performance of the original classifier. Under the setting of 30 samples in the endangered species dataset, the average detection accuracy value of the method, mAP50, reaches 76.54%, which is 7.98% higher than that of the pre-improved FSCE method. This paper also compares the algorithm on the PASCOL VOC dataset, which is optimal and has good generalization ability compared to the other five algorithms.

Application of an improved VGG and RPN network in precision parts recognition

Although various automatic or semi-automatic recognition algorithms have been proposed for tiny part recognition, most of them are limited to expert knowledge base-based target recognition techniques, which have high false detection rates, low recognition accuracy and low efficiency, which largely limit the quality as well as efficiency of tiny part assembly. Therefore, this paper proposes a precision part image preprocessing method based on histogram equalization algorithm and an improved convolutional neural network (i.e. Region Proposal Network(RPN), Visual Geometry Group(VGG)) model for precision recognition of tiny parts. Firstly, the image is restricted to adaptive histogram equalization for the problem of poor contrast between part features and the image background. Second, a custom central loss function is added to the recommended frame extraction RPN network to reduce problems such as excessive intra-class spacing during classification. Finally, the local response normalization function is added after the nonlinear activation function and pooling layer in the VGG network, and the original activation function is replaced by the Relu function to overcome the problems such as high nonlinearity and serious overfitting of the original model. Experiments show that the improved VGG model achieves 95.8% accuracy in precision part recognition and has a faster recognition speed than most existing convolutional networks trained on the same test set.

Detection of Residual Film on the Field Surface Based on Faster R-CNN Multiscale Feature Fusion

After the residual film recycling machine recovers the film, some small pieces of the film will remain on the surface of the field. To solve the problem of collecting small pieces of film, it is necessary to develop a piece of intelligent picking equipment. The detection of small pieces of film is the first problem to be solved. This study proposes a method of an object detection algorithm fusing multi-scale features (MFFM Faster R-CNN) based on improved Faster R-CNN. Based on the Faster R-CNN model, the feature pyramid network is added to solve the problem of multiscale change of residual film. The convolution block attention module is introduced to enhance the feature extraction ability of the model. The Soft-NMS algorithm is used instead of the NMS algorithm to improve the detection accuracy of the model in the RPN network. The experimental results show that the model is able to effectively detect surface residual film in complex environments, with an AP of 83.45%, F1-score of 0.89, and average detection time of 248.36 ms. The model is compared with SSD and YOLOv5 under the same experimental environment and parameters, and it is found that the model not only ensures high-precision detection but also ensures real-time detection. This lays the theoretical foundation for the subsequent development of field surface residual film intelligent picking equipment.

Method for Detection of Ripe Navel Orange Fruit on Trees in Various Weather

<p>The algorithm used is based on the Faster R-CNN network. It uses resnet101_vd as the backbone network to extract navel orange image features. It minimizes the error between the inferred bounding box and the actual labeled bounding box through the RPN network and the ROI Pooling layer and non-maximum suppression method. The AP during model training reached 92.34% on sunny days, 96.84% on cloudy days, and 90.05% on foggy days. When evaluating the model, it reached 92.34% on sunny days, 96.89% on cloudy days, and 89.2% on foggy days. The processing time of this model is 63.84fps on sunny days, 68.6fps on cloudy days, and 56.29fps on foggy days. It meets the requirements of rapid and accurate identification in actual picking. In addition, it compares this model with the Faster RCNN with vgg16 as the backbone network and YOLO-v4 models. It effectively improves detection accuracy and speed. Moreover, it reduces the number of false detection and missed detection of navel orange detection. It im-proves position accuracy. This paper realizes the efficient detection of ripe navel orange fruits on trees in various weather. </p> <p> </p>

An approach for characterization of infected area in tomato leaf disease based on deep learning and object detection technique

Tomato leaf infections are a common threat to long-term tomato production that affects many farmers worldwide. Early detection, treatment, and solution of tomato leaf specificity are critical for promoting healthy tomato plant growth and ensuring ample supply and health security for the world’s geometric growth (population). The detection of plant leaf disease using computer-assisted technologies is prevalent these days. In this work, use the 1610 tomato leaf images of different classes from PlantVillage standard repository for the localization of objects. An effective Deep Learning (DL) modified Mask Region Convolutional Neural Network (Mask R-CNN) is proposed for the autonomous segmentation and detection of tomato plant leaf disease in this research. Intending to conserve memory space and computational expense, the suggested model adds a light head “Region Convolutional Neural Network (R-CNN)”. By varying the proportions of anchor in the RPN network and also changing the feature extraction topology, which improves the detection accuracy and computing the metric performance. The proposed technique is compared to existing state-of-the-art models to check if it is viable and robust. The outcomes of the suggested model achieved the results in terms of Mean Average Precision (mAP), F1-score, and accuracy of 0.88, 0.912, and 0.98, respectively. Furthermore, as the model’s ability increases with some parameters, the detection time for lesion detection is reduced by two times than the existing models.

Multiple Diseases and Pests Detection Based on Federated Learning and Improved Faster R-CNN

Traditional disease and pest detection technology employ cloud based deep learning, which facing the pressures such as high data storage and communication costs, unbalanced and insufficient data from orchards, diversity of pests and diseases, and complex detection environments. In this paper, we propose a multiple pest detection technique based on Federated Learning (FL) and improved Faster Region Convolutional Neural Network (R-CNN). As the new distributed computing model, FL can derive a shared model integrating the advantages of data from all parties without uploading local data, and also reduces the communication cost. A restriction M is added to the FL algorithm to ensure the convergence of the model and improve the training speed. According to the original Faster R-CNN network, ResNet-101 is used instead of VGG-16 to construct the base convolutional layer to maintain the original structure of small-sized targets and improve the detection speed. Then, the multi-size fusion of feature maps from different convolutional layers is performed to improve the detection accuracy of multi-size multiple pests and diseases. Finally, a Soft-NMS algorithm is proposed to solve the apple obscured problem after the RPN network. The experimental results show that the improved Faster R-CNN can achieve an average accuracy of 90.27% on multiple pest detection, and the detection time is only 0.05 seconds per image. After using FL, the mAP of the model reached 89.34% and the model training speed was improved by 59%.

Real-Time Leaf Recognition Method Based on Image Segmentation and Feature Extraction

Leaf recognition has been an important research field of image recognition in the recent past. However, traditional leaf recognition methods can be easily affected by environments and cannot realize multi-leaf recognition under a complex background in real time. In this work, we present a real-time leaf recognition method based on image segmentation and feature recognition. First, we denoise the input of a leaf image, performing a leaf segmentation with an improved FCN network model, and then optimize the contour edge with a CRF algorithm to get a leaf segmentation image. Second, we extract the content features of the segmented leaf image with an Inception-V2 network model to get a feature map of the leaf image. Third, we input the feature map into an RPN network to obtain a set of regional candidate frames and then integrate the feature map and the information of candidate frames in a RoI Pooling layer, which can extract the feature map of a candidate frame area and scale it to a fixed-size feature map. Finally, we send the feature map to a fully connected layer to classify each preselection box content through the calculation of preselection feature maps, and then obtain the final accurate position of the prediction box by utilizing a bounding box regression. The experimental results show that the proposed method can achieve multi-leaf recognitions with high accuracy and fast speed under complex environments in real time.

Face mask detection algorithm for multi task recognition in complex environment

The right wearing masks can effectively prevent the spread of viruses. There are usually complex distractions in densely populated environments, which will affect detection tasks of face masks. However, the current detecting algorithm based on Faster R-CNN can not meet mask wearing detection of small targets in complex environments, so this paper puts forward a mask wearing detection algorithm based on the improved Faster R-CNN. This algorithm changes the traditional single RPN network model into one multi-task enhanced RPN model, which can improve the detection and recognition accuracy, the improved Soft-NMS algorithm is used to delete the redundant candidate boxes of regional candidate network output. At the same time, CIoU loss function with better performance, which fully considers the center point distance between the target and the detection frame and improves the accuracy of detection, replaced the original IoU loss function. 3000 sample images are used to do training and experimental verification of the model according to 1:3ratio of positive and negative samples. Experimental results show that compared with the Faster R-CNN algorithm, the algorithm proposed in this paper improves the face detection accuracy and face mask wearing detection by 7.15% and 15.99%respectively. Besides, the average detection accuracy improved by 11.57% and the FPS by 4.6.

Multi label image retrieval based on depth hash

In recent years, hash is a popular method of image retrieval. Convolutional neural network is used to generate image features and hash codes, so as to achieve fast and effective image retrieval. In order to solve the problem of poor image retrieval effect caused by complex background of Multi-target Image, this paper proposes a hash generation method based on the combination of target region recommendation network and convolutional neural network. The RPN network is selected as the target region recommendation algorithm. After the image passes through the RPN network, multiple target regions and the four-dimensional coordinates of each region will be generated. According to the four-dimensional coordinates, the main target region will be filtered and extracted. Then, the feature of the largest target area is extracted by the convolutional neural network (GoogleNet), and the corresponding hash code is generated, which is finally retrieved in the database. Experiments are carried out on voc2012 data set and self collected data set to verify the algorithm. When the number of test images is 1000, the experimental results show that the total correct rate of the retrieval results of this method is 95.5%, which is about 5 percentage points higher than the existing methods.

Microscopic Tumour Classification by Digital Mammography.

In this paper, we investigate the classification of microscopic tumours using full digital mammography images. Firstly, to address the shortcomings of traditional image segmentation methods, two different deep learning methods are designed to achieve the segmentation of uterine fibroids. The deep lab model is used to optimize the lesion edge detailed information by using the void convolution algorithm and fully connected CRF, and the two semantic segmentation networks are compared to obtain the best results. The Mask RCNN case segmentation model is used to effectively extract features through the ResNet structure, combined with the RPN network to achieve effective use and fusion of features, and continuously optimize the network training to achieve a fine segmentation of the lesion area, and demonstrate the accuracy and feasibility of the two models in medical image segmentation. Histopathology was used to obtain ER, PR, HER scores, and Ki-67 percentage values for all patients. The Kaplan-Meier method was used for survival estimation, the Log-rank test was used for single-factor analysis, and Cox proportional risk regression was used for multifactor analysis. The prognostic value of each factor was calculated, as well as the factors affecting progression-free survival. This study was done to compare the imaging characteristics and diagnostic value of mammography and colour Doppler ultrasonography in nonspecific mastitis, improve the understanding of the imaging characteristics of nonspecific mastitis in these two examinations, improve the accuracy of the diagnosis of this type of disease, improve the ability of distinguishing it from breast cancer, and reduce the rate of misdiagnosis.

Crack detection based on multi-scale Faster RCNN with attention

The background of the EL image of a photovoltaic cell under electroluminescence (EL) presents complex non-uniform texture features, and there are grain pseudo-defects similar to cracks. At the same time, the cracks appear as multi-scale features with various shapes. The above mentioned difficulties have presented great chal-lenges for the detection task. Therefore, this paper proposes a multi-scale Faster-RCNN model that integrates at-tention. On the one hand, an improved feature pyramid network is used to obtain multi-scale advanced semantic feature maps to improve the network's feature expression ability of multi-scale crack defects. On the other hand, an improved attention region proposal network A-RPN is adopted to increase the model's attention to crack defects and suppress the characteristics of complex background and grain pseudo-defects. At the same time, in the RPN network training process, a loss function Focal loss is used to reduce the proportion of simple samples in the training process, so that the model pays more attention to the samples that are difficult to distinguish. Experimental results show that this algorithm improves the accuracy of crack defect detection in EL images, reaching nearly 95%.

News Video Title Extraction Algorithm Based on Deep Learning

In order to better serve the needs of news business, researchers apply the information extraction technology of news headlines to the news field to assist decision-making. This article designs the shared convolutional layer and RPN network in the detection network respectively, and improves the depth of the shared convolutional layer, using VGGNet instead of ZFNet. A classification method incorporating semantic enhancement is designed for fine-grained topic category news classification. Combined with the idea of modular fusion mechanism, a semantic-enhanced classification model Multiple-Fusion is proposed. The Bert module replaces the traditional Word2Vec for semantic vector characterization, and introduces Bi-LSTM to adaptively extract context features, strengthens feature expression through the self-attention mechanism and adjusts network weights, and finally makes the model achieve accurate classification. This article designs a novel word-level data enhancement strategy for text data enhancement, which solves the problems of fewer training corpus samples and model overfitting. This article proposes a video target extraction method based on deep learning. The algorithm framework includes particle filtering, pre-training features, convolutional neural networks, discriminative classifiers, and online parameter updates. This method deeply combines deep learning models and traditional target extraction methods and frameworks. The experimental results show that the method in this article has relatively outstanding performance, and can adapt to many interferences encountered in the extraction process and the change of the target itself, and has strong robustness.

A Traffic-Sign Detection Algorithm Based on Improved Sparse R-cnn

Automatic traffic-sign detection is a hot topic in computer vision and one of the critical technologies of intelligent transportation. The Transformer structure has recently become a research hotspot due to its excellent performance. We hope to apply this structure to the design of traffic sign detection algorithms. Therefore, we make some improvements to Sparse R-cnn, a neural network model inspired by Transformer. Sparse R-cnn is a novel model, and its core idea is to replace hundreds of thousands of candidate anchors in the RPN network with a small set of proposal boxes. The experiments in our paper proved that the performance of the Sparse R-cnn model is better than other existing general object detection models. Based on the original Sparse R-cnn inspiration, an improved Sparse R-cnn model is proposed. First, a novel backbone for the task of traffic-sign detection is proposed. Multi-scale fusion structure is the essential method of improving the algorithm for small target detection, so improving the multi-scale capability of the backbone is a required method for designing traffic sign detection. So, we made further improvements to the existing backbone ResNest. We enhanced the multi-scale representation ability of the backbone by constructing hierarchical residual-like connections within each single radix block in the original ResNest. We call the improved backbone Res2Nest. The novel backbone proposed by us shows better performance without introducing excessive computational costs to the model. In addition, the attention mechanism is also an effective method to improve the detection of traffic signs, so we set up a branch network for recalibrating the channel feature response adaptively through the Global Average Pooling (GAP) operation and a fully connected layer. It can also be seen as the implementation of the cross-channel self-attention mechanism. After experiments by TT100K dataset, our method would attain a better accuracy and robustness.

Research on image segmentation of inner cylinder wall with annular weld based on deep learning

Based on the characteristics of autonomous learning with deep learning and high recognition rate, an image segmentation study based on Faster R-CNN for the inner wall annular weld is proposed. The method extracts pixels containing annular weld feature information by using an RPN network to improve target detection speed. Then, the convolutional layer output detection model is shared by the RPN network and the Fast R-CNN network to realize accurate detection of the annular weld in the video image. The research results show that the proposed method can accurately detect the circumferential weld bead and segment the image under the condition of poor video image quality, and has the advantages of strong anti-interference ability and accurate identification.

Research on Detection of Large Coal Blockage at the Transfer Point of Belt Conveyor Based on Improved Mask R-CNN

Aiming at the problem of coal accumulation and blockage at the transfer point at the transfer point due to the presence of large coal in the coal belt conveyor, a method based on the improved Mask R-CNN for the detection of the jam at the transfer point of the belt conveyor is proposed. The method firstly adds the SENET and SKNet models to the Mask R-CNN feature extraction part ResNet50, enhances the function of extracting features, and cancels the segmentation part of Mask R-CNN. In order to make the network better converge, the loss function in the RPN network is optimized, and the original IOU is replaced with GIOU, which solves the problem that the IOU cannot be optimized when it is 0 and the IOU cannot distinguish the anchor from the ground truth.

Northern Maize Leaf Blight Detection Under Complex Field Environment Based on Deep Learning

Northern maize leaf blight is one of the major diseases that endanger the health of maize. The complex background of the field and different light intensity make the detection of diseases more difficult. A multi-scale feature fusion instance detection method, based on convolutional neural network, is proposed to detect maize leaf blight. The proposed technique incorporates three major steps of data set preprocessing part, fine-tuning network and detection module. In the first step, the improved retinex is used to process data sets, which successfully solves the problem of poor detection effects caused by high-intensity light. In the second step, the improved RPN is utilized to adjust the anchor box of diseased leaves. The improved RPN network identifies and deletes negative anchors, which reduces the search space of the classifier and provides better initial information for the detection network. In this paper, a transmission module is designed to connect the fine-tuning network with the detection module. On the one hand, the transmission module fuses the features of the low-level and high-level to improve the detection accuracy of small target diseases. On the other hand, the transmission module converts the feature map associated with the fine-tuning network to the detection module, thus realizing the feature sharing between the detection module and the fine-tuning network. In the third step, the detection module takes the optimized anchor as input, focuses on detecting the diseased leaves. By sharing the features of the transmission module, the time-consuming process of using candidate regions layer by layer to detect is eliminated. Therefore, the efficiency of the whole model has reached the efficiency of the one-stage model. In order to further optimize the detection effect of the model, we replace the loss function with generalized intersection over union (GIoU). After 60000 iterations, the highest mean average precision (mAP) reaches 91.83%. The experimental results indicate that the improved model outperforms several existing methods in terms of greater precision and frames per second (FPS).

Cross-scale fusion detection with global attribute for dense captioning

As a new task of image understanding, the dense caption model needs to locate and describe a salient region in the image simultaneously. It inevitably divides the dense caption model into two parts, one part for detecting the regions of interest and the other part for generating regional language caption. Previous methods are relatively simple to deal with these two parts, using the feature map on the last convolution layer of RPN network to predict object coordinates, and using LSTM for regional language modeling. However, the structure of RPN is insufficient to deal with a large number of objects in the complex dataset, and LSTM also fails to effectively utilize the global information of images in regional language training, which brings opportunity to improve the performance of dense caption. In this paper, we propose a novel Cross-scale Fusion with Global Attribute model (CSGA) that enables the two parts of the dense caption model to perform normal end-to-end training without mutual interference. Furthermore, our model uses a one-stage object detector with feature map fusion operation across multiple detection scales to improve the quality of object detection part, and combines image features with the global high-level attribute to improve regional language training. We design a variety of model architectures and conducted sufficient experiments. Empirical results on Visual Genome dataset show that our model achieves competitive results with mAP 8.33.

Research on Pipeline Defect Detection Based on Optimized Faster R-cnn Algorithm

Aiming at the problem of low efficiency and high labor intensity in manual inspection of drainage pipes, a method of defect inspection of drainage pipes based on optimized Faster-Rcnn algorithm is proposed. Faster R-CNN is an algorithm proposed for target detection this year, which is based on the deep learning network model of region-based recommendation network (RPN). This paper analyzes the implementation of RPN network in Faster R-CNN algorithm, optimizes the network and introduces K-Means clustering method. By clustering all Anchors in the training set and inputting the clustering results into the RPN network, the training of the network can be accelerated and the recognition accuracy of the algorithm can be improved. The experimental results show that the accuracy of the algorithm is up to 92.4%, which has great application value. This research has important reference significance to promote the automatic detection of sewage pipeline defects.