Improvement In Average Precision Research Articles

Pedestrian Detection Based on Panoramic Depth Map Transformed from 3D-LiDAR Data

Object detection is a crucial task of autonomous driving. This paper addresses an effective algorithm for pedestrian detection of the panoramic depth map transformed from the 3D-LiDAR data. Firstly, the 3D point clouds are transformed into panoramic depth maps, and then the panoramic depth maps are enhanced. Secondly, the grounds of the 3D point clouds are removed. The remaining point clouds are clustered, filtered and projected onto the previously generated panoramic depth maps, and new panoramic depth maps are obtained. Finally, the new panoramic depth maps are jointed to generate depth maps with different sizes, which are used as input of the improved PVANET for pedestrian detection. The 2D image of the panoramic depth map applied to the proposed algorithm is transformed from 3D point cloud, effectively containing the panorama of the sensor, and is more suitable for the environment perception of autonomous driving. Compared with the detection algorithm based on RGB images, the proposed algorithm cannot be affected by light, and can maintain the normal average precision of pedestrian detection at night. In order to increase the robustness of detecting small objects like pedestrians, the network structure based on the original PVANET is modified in this paper. A new dataset is built by processing the 3D-LiDAR data and the model trained on the new dataset perform well. The experimental results show that the proposed algorithm achieves high accuracy and robustness in pedestrian detection under different illumination conditions. Furthermore, when trained on the new dataset, the model exhibits average precision improvements of 2.8–5.1 % over the original PVANET, making it more suitable for autonomous driving applications.

IoT with Cloud Centric Vehicle Detection and Counting System for Smart Traffic Surveillance

The efficient management of road traffic is one primary facet of many, in smart cities. Traffic overcrowding can be managed successfully, if prior estimation of the number of vehicles that will pass though a crowded junction in a specific time is known. This paper introduces a methodology which targets vehicle extraction on videos covering vehicles. To resolve the problem of current vehicle detection such as the need of detection accuracy and slow speed, an improved YOLOv3 vehicle detection is utilized. The k-means clustering used to group the bounding box around the vehicle in training dataset. The method for calculation of loss with respect to the length and width of the bounding boxes was recovered through the implementation of the batch normalization process. Finally, to improve the feature extraction of the network the high repeated convolution layer are removed. The experiment results are carried out on the BIT-vehicle validation datasets which shows the improvement of mean Average Precision (mAP) could certainly reach 95.6%.

EOVNet: Earth-Observation Image-Based Vehicle Detection Network

Vehicle detection from earth-observation (EO) image has been attracting remarkable attention for its critical value in a variety of applications. Encouraged by the stunning success of deep learning techniques based on convolutional neural networks (CNNs), which have revolutionized the visual data processing community and obtained the state-of-the-art performance in a variety of classification and recognition tasks on benchmark datasets, we propose a network, called EOVNet (EO image-based vehicle detection network), to bridge the gap between the advanced deep learning research of object detection and the specific task of vehicle detection in EO images. Our network has integrated nearly all advanced techniques including very deep residual networks for feature extraction, feature pyramid to fuse multiscale features, network for proposal generation with feature sharing, and hard example mining. Moreover, our novel designs of probability-based localization and homography-based data augmentation have been investigated, resulting in further improvement of the detection performance. For performance evaluation, we have collected nearly all the representative EO datasets associated with vehicle detection. Extensive experiments on the representative datasets demonstrate that our method outperforms the state-of-the-art object detection approach Faster R-CNN++ (which is based on the Faster R-CNN framework, but with significant improvement) with 5% average precision improvement. The source code will be made available after the review process.

Heterogeneous Gray-Temperature Fusion-Based Deep Learning Architecture for Far Infrared Small Target Detection

This paper proposes the end-to-end detection of a deep network for far infrared small target detection. The problem of detecting small targets has been a subject of research for decades and has been applied mainly in the field of surveillance. Traditional methods focus on filter design for each environment, and several steps are needed to obtain the final detection result. Most of them work well in a given environment but are vulnerable to severe clutter or environmental changes. This paper proposes a novel deep learning-based far infrared small target detection method and a heterogeneous data fusion method to solve the lack of semantic information due to the small target size. Heterogeneous data consists of radiometric temperature data (14-bit) and gray scale data (8-bit), which includes the physical meaning of the target, and compares the effects of the normalization method to fuse heterogeneous data. Experiments were conducted using an infrared small target dataset built directly on the cloud backgrounds. The experimental results showed that there is a significant difference in performance according to the various fusion methods and normalization methods, and the proposed detector showed approximately 20% improvement in average precision (AP) compared to the baseline constant false alarm rate (CFAR) detector.

G -Softmax: Improving Intraclass Compactness and Interclass Separability of Features.

Intraclass compactness and interclass separability are crucial indicators to measure the effectiveness of a model to produce discriminative features, where intraclass compactness indicates how close the features with the same label are to each other and interclass separability indicates how far away the features with different labels are. In this paper, we investigate intraclass compactness and interclass separability of features learned by convolutional networks and propose a Gaussian-based softmax ( G -softmax) function that can effectively improve intraclass compactness and interclass separability. The proposed function is simple to implement and can easily replace the softmax function. We evaluate the proposed G -softmax function on classification data sets (i.e., CIFAR-10, CIFAR-100, and Tiny ImageNet) and on multilabel classification data sets (i.e., MS COCO and NUS-WIDE). The experimental results show that the proposed G -softmax function improves the state-of-the-art models across all evaluated data sets. In addition, the analysis of the intraclass compactness and interclass separability demonstrates the advantages of the proposed function over the softmax function, which is consistent with the performance improvement. More importantly, we observe that high intraclass compactness and interclass separability are linearly correlated with average precision on MS COCO and NUS-WIDE. This implies that the improvement of intraclass compactness and interclass separability would lead to the improvement of average precision.

Query-by-Example Speech Search Using Recurrent Neural Acoustic Word Embeddings With Temporal Context

Acoustic word embeddings (AWEs) have been popular in low-resource query-by-example speech search. They are using vector distances to find the spoken query in search content, which has much lower computation than the conventional dynamic time warping (DTW)-based approaches. The AWE networks are usually trained using variable-length isolated spoken words, while they are applied to fixed-length speech segments obtained by shifting an analysis window on speech content. There is an obvious mismatch between the learning of AWEs and its application on search content. To mitigate such mismatch, we propose to include temporal context information on spoken word pairs to learn recurrent neural AWEs. More specifically, the spoken word pairs are represented by multi-lingual bottleneck features (BNFs) and padded with the neighboring frames of the target spoken words to form fixed-length speech segment pairs. A deep bidirectional long short-term memory (BLSTM) network is then trained with a triplet loss using the speech segment pairs. Recurrent neural AWEs are obtained by concatenating the BLSTM backward and forward outputs. During QbE speech search stage, both spoken query and search content are converted into recurrent neural AWEs. Cosine distances are then measured between them to find the spoken query. The experiments show that using temporal context is essential to alleviate the mismatch. The proposed recurrent neural AWEs trained with temporal context outperform the previous state-of-art features with 12.5% relative mean average precision (MAP) improvement on QbE speech search.

Application of hybrid of Fuzzy Set, Trust and Genetic Algorithm in query log mining for effective Information Retrieval

The precision of Information Retrieval (IR) System is low due to imprecise user queries as well as because of information overload on web. The Fuzzy set infers the user’s information need from vague and imprecise queries and web recommender systems are used to overcome information overload problem. The performance of recommender system is still low due to data sparsity. The concept of trust is used to deal with data sparseness problem and improves the performance of recommender system. Optimization techniques like Genetic Algorithm(GA) have been applied in domain of information retrieval for effective web search. In this research hybrid of Fuzzy set, GA and Trust has been used together in query log mining for personalized web search based on using fuzzy queries for recommendation of optimal set of trusted documents. Thus the use of hybrid of Fuzzy set, trust and GA together infer the user’s information need from vague and imprecise user’s queries and optimize the web page ranking of trusted web pages for effective personalized web search. The experimental results were analyzed statistically as well as compared with GA IR, and Fuzzy Trust based IR. Hence based on comparative analysis of results, thus hybrid of Fuzzy Set, Trust and GA shows the improvement in average precision of search results and confirms the effective personalization of web search.

Random Access Memories: A New Paradigm for Target Detection in High Resolution Aerial Remote Sensing Images.

We propose a new paradigm for target detection in high resolution aerial remote sensing images under small target priors. Previous remote sensing target detection methods frame the detection as learning of detection model + inference of class-label and bounding-box coordinates. Instead, we formulate it from a Bayesian view that at inference stage, the detection model is adaptively updated to maximize its posterior that is determined by both training and observation. We call this paradigm "random access memories (RAM)." In this paradigm, "Memories" can be interpreted as any model distribution learned from training data and "random access" means accessing memories and randomly adjusting the model at detection phase to obtain better adaptivity to any unseen distribution of test data. By leveraging some latest detection techniques e.g., deep Convolutional Neural Networks and multi-scale anchors, experimental results on a public remote sensing target detection data set show our method outperforms several other state of the art methods. We also introduce a new data set "LEarning, VIsion and Remote sensing laboratory (LEVIR)", which is one order of magnitude larger than other data sets of this field. LEVIR consists of a large set of Google Earth images, with over 22 k images and 10 k independently labeled targets. RAM gives noticeable upgrade of accuracy (an mean average precision improvement of 1% ~ 4%) of our baseline detectors with acceptable computational overhead.

Assessment of contact predictions in CASP12: Co-evolution and deep learning coming of age.

Following up on the encouraging results of residue‐residue contact prediction in the CASP11 experiment, we present the analysis of predictions submitted for CASP12. The submissions include predictions of 34 groups for 38 domains classified as free modeling targets which are not accessible to homology‐based modeling due to a lack of structural templates. CASP11 saw a rise of coevolution‐based methods outperforming other approaches. The improvement of these methods coupled to machine learning and sequence database growth are most likely the main driver for a significant improvement in average precision from 27% in CASP11 to 47% in CASP12. In more than half of the targets, especially those with many homologous sequences accessible, precisions above 90% were achieved with the best predictors reaching a precision of 100% in some cases. We furthermore tested the impact of using these contacts as restraints in ab initio modeling of 14 single‐domain free modeling targets using Rosetta. Adding contacts to the Rosetta calculations resulted in improvements of up to 26% in GDT_TS within the top five structures.

A Wikipedia powered state-based approach to automatic search query enhancement

This paper describes the development and testing of a novel Automatic Search Query Enhancement (ASQE) algorithm, the Wikipedia N Sub-state Algorithm (WNSSA), which utilises Wikipedia as the sole data source for prior knowledge. This algorithm is built upon the concept of iterative states and sub-states, harnessing the power of Wikipedia’s data set and link information to identify and utilise reoccurring terms to aid term selection and weighting during enhancement. This algorithm is designed to prevent query drift by making callbacks to the user’s original search intent by persisting the original query between internal states with additional selected enhancement terms. The developed algorithm has shown to improve both short and long queries by providing a better understanding of the query and available data. The proposed algorithm was compared against five existing ASQE algorithms that utilise Wikipedia as the sole data source, showing an average Mean Average Precision (MAP) improvement of 0.273 over the tested existing ASQE algorithms.

Membrane protein contact and structure prediction using co-evolution in conjunction with machine learning.

De novo membrane protein structure prediction is limited to small proteins due to the conformational search space quickly expanding with length. Long-range contacts (24+ amino acid separation)–residue positions distant in sequence, but in close proximity in the structure, are arguably the most effective way to restrict this conformational space. Inverse methods for co-evolutionary analysis predict a global set of position-pair couplings that best explain the observed amino acid co-occurrences, thus distinguishing between evolutionarily explained co-variances and these arising from spurious transitive effects. Here, we show that applying machine learning approaches and custom descriptors improves evolutionary contact prediction accuracy, resulting in improvement of average precision by 6 percentage points for the top 1L non-local contacts. Further, we demonstrate that predicted contacts improve protein folding with BCL::Fold. The mean RMSD100 metric for the top 10 models folded was reduced by an average of 2 Å for a benchmark of 25 membrane proteins.

Characteristics of the BDS Carrier Phase Multipath and Its Mitigation Methods in Relative Positioning.

The carrier phase multipath effect is one of the most significant error sources in the precise positioning of BeiDou Navigation Satellite System (BDS). We analyzed the characteristics of BDS multipath, and found the multipath errors of geostationary earth orbit (GEO) satellite signals are systematic, whereas those of inclined geosynchronous orbit (IGSO) or medium earth orbit (MEO) satellites are both systematic and random. The modified multipath mitigation methods, including sidereal filtering algorithm and multipath hemispherical map (MHM) model, were used to improve BDS dynamic deformation monitoring. The results indicate that the sidereal filtering methods can reduce the root mean square (RMS) of positioning errors in the east, north and vertical coordinate directions by 15%, 37%, 25% and 18%, 51%, 27% in the coordinate and observation domains, respectively. By contrast, the MHM method can reduce the RMS by 22%, 52% and 27% on average. In addition, the BDS multipath errors in static baseline solutions are a few centimeters in multipath-rich environments, which is different from that of Global Positioning System (GPS) multipath. Therefore, we add a parameter representing the GEO multipath error in observation equation to the adjustment model to improve the precision of BDS static baseline solutions. And the results show that the modified model can achieve an average precision improvement of 82%, 54% and 68% in the east, north and up coordinate directions, respectively.

L(p) -norm IDF for scalable image retrieval.

The inverse document frequency (IDF) is prevalently utilized in the bag-of-words-based image retrieval application. The basic idea is to assign less weight to terms with high frequency, and vice versa. However, in the conventional IDF routine, the estimation of visual word frequency is coarse and heuristic. Therefore, its effectiveness is largely compromised and far from optimal. To address this problem, this paper introduces a novel IDF family by the use of Lp-norm pooling technique. Carefully designed, the proposed IDF considers the term frequency, document frequency, the complexity of images, as well as the codebook information. We further propose a parameter tuning strategy, which helps to produce optimal balancing between TF and pIDF weights, yielding the so-called Lp-norm IDF (pIDF). We show that the conventional IDF is a special case of our generalized version, and two novel IDFs, i.e., the average IDF and the max IDF, can be defined from the concept of pIDF. Further, by counting for the term-frequency in each image, the proposed pIDF helps to alleviate the visual word burstiness phenomenon. Our method is evaluated through extensive experiments on four benchmark data sets (Oxford 5K, Paris 6K, Holidays, and Ukbench). We show that the pIDF works well on large scale databases and when the codebook is trained on irrelevant data. We report an mean average precision improvement of as large as +13.0% over the baseline TF-IDF approach on a 1M data set. In addition, the pIDF has a wide application scope varying from buildings to general objects and scenes. When combined with postprocessing steps, we achieve competitive results compared with the state-of-the-art methods. In addition, since the pIDF is computed offline, no extra computation or memory cost is introduced to the system at all.

MRI and CT image indexing and retrieval using local mesh peak valley edge patterns

In this paper, a new pattern based feature, local mesh peak valley edge pattern (LMePVEP) is proposed for biomedical image indexing and retrieval. The standard LBP extracts the gray scale relationship between the center pixel and its surrounding neighbors in an image. Whereas the proposed method extracts the gray scale relationship among the neighbors for a given center pixel in an image. The relations among the neighbors are peak/valley edges which are obtained by performing the first-order derivative. The performance of the proposed method (LMePVEP) is tested by conducting two experiments on two benchmark biomedical databases. Further, it is mentioned that the databases used for experiments are OASIS−MRI database which is the magnetic resonance imaging (MRI) database and VIA/I–ELCAP-CT database which includes region of interest computer tomography (CT) images. The results after being investigated show a significant improvement in terms average retrieval precision (ARP) and average retrieval rate (ARR) as compared to LBP and LBP variant features.

Model-Based Unsupervised Spoken Term Detection with Spoken Queries

We present a set of model-based approaches for unsupervised spoken term detection (STD) with spoken queries that requires neither speech recognition nor annotated data. This work shows the possibilities in migrating from DTW-based to model-based approaches for unsupervised STD. The proposed approach consists of three components: self-organizing models, query matching, and query modeling. To construct the self-organizing models, repeated patterns are captured and modeled using acoustic segment models (ASMs). In the query matching phase, a document state matching (DSM) approach is proposed to represent documents as ASM sequences, which are matched to the query frames. In this way, not only do the ASMs better model the signal distributions and time trajectories of speech, but the much-smaller number of states than frames for the documents leads to a much lower computational load. A novel duration-constrained Viterbi (DC-Vite) algorithm is further proposed for the above matching process to handle the speaking rate distortion problem. In the query modeling phase, a pseudo likelihood ratio (PLR) approach is proposed in the pseudo relevance feedback (PRF) framework. A likelihood ratio evaluated with query/anti-query HMMs trained with pseudo relevant/irrelevant examples is used to verify the detected spoken term hypotheses. The proposed framework demonstrates the usefulness of ASMs for STD in zero-resource settings and the potential of an instantly responding STD system using ASM indexing. The best performance is achieved by integrating DTW-based approaches into the rescoring steps in the proposed framework. Experimental results show an absolute 14.2% of mean average precision improvement with 77% CPU time reduction compared with the segmental DTW approach on a Mandarin broadcast news corpus. Consistent improvements were found on TIMIT and MediaEval 2011 Spoken Web Search corpus.

Automated query learning with Wikipedia and genetic programming

Most of the existing information retrieval systems are based on bag-of-words model and are not equipped with common world knowledge. Work has been done towards improving the efficiency of such systems by using intelligent algorithms to generate search queries, however, not much research has been done in the direction of incorporating human-and-society level knowledge in the queries. This paper is one of the first attempts where such information is incorporated into the search queries using Wikipedia semantics. The paper presents Wikipedia-based Evolutionary Semantics (Wiki-ES) framework for generating concept based queries using a set of relevance statements provided by the user. The query learning is handled by a co-evolving genetic programming procedure.To evaluate the proposed framework, the system is compared to a bag-of-words based genetic programming framework as well as to a number of alternative document filtering techniques. The results obtained using Reuters newswire documents are encouraging. In particular, the injection of Wikipedia semantics into a GP-algorithm leads to improvement in average recall and precision, when compared to a similar system without human knowledge. A further comparison against other document filtering frameworks suggests that the proposed GP-method also performs well when compared with systems that do not rely on query-expression learning.

Image retrieval using improved texton co-occurrence matrix

This paper proposed a new technique for content-based image retrieval (CBIR) using a combination of a` trous wavelet transform (AWT) and Julesz|s texton elements. AWT is used to decomposse the image into different scales and different texton elements are used to detect the spatial co-relation among the transform coefficients in horizontal, vertical, diagonal and minor diagonal directions which results into texton image. Further co-occurance operations are performed on texton image in horizontal and zigzag direction. Texton co-occurance matrix gives second order statistics related to texton image. The proposed method is tested on natural database (Corel 1000 and Corel 2500) and texture database (Brodatz and USC texture database) and the retrieval results have demonstrated significant improvement in average precision, average recall rate, as well as feature extraction and retrieval time compared to optimal quantised wavelet correlogram (OQWC) and Gabor wavelet correlogram (GWC).

Performance Analysis for Lattice-Based Speech Indexing Approaches Using Words and Subword Units

Lattice-based speech indexing approaches are attractive for the combination of short spoken segments, short queries, and low automatic speech recognition (ASR) accuracies, as lattices provide recognition alternatives and therefore tend to compensate for recognition errors. Position-specific posterior lattices (PSPLs) and confusion networks (CNs), two of the most popular lattice-based approaches, both reduce disk space requirements and are more efficient than raw lattices. When PSPLs and CNs are used in a word-based fashion, they cannot handle OOV or rare word queries. In this paper, we propose an efficient approach for the construction of subword-based PSPLs (S-PSPLs) and CNs (S-CNs) and present a comprehensive performance analysis of PSPL and CN structures using both words and subword units, taking into account basic principles and structures, and supported by experimental results on Mandarin Chinese. S-PSPLs and S-CNs are shown to yield significant mean average precision (MAP) improvements over word-based PSPLs and CNs for both out-of-vocabulary (OOV) and in-vocabulary queries while requiring much less disk space for indexing.

A new feature for image retrieval using a`trous wavelet transform and textons

In this paper, a combination of a`trous wavelet transform (AWT) and Julesz's texton theory is used for feature extraction and retrieval of the images from natural image database. AWT is used to decompose the image and different texton elements are used to detect the spatial co-relation among the transform pixels in horizontal, vertical, diagonal and minor diagonal directions. Further, this information is combined with texture oriented image for generation of image feature vector. The proposed method is tested on Corel 1000 and 2500 image database and the retrieval results have demonstrated significant improvement in average precision, average weighted precision, average recall rate, average rank, standard deviation of rank, standard deviation of precision as well as feature extraction and retrieval time compared to optimal quantised wavelet correlogram (OQWC) and Gabor wavelet correlogram (GWC).

Preparing heterogeneous XML for full-text search

XML retrieval is facing new challenges when applied to heterogeneous XML documents, where next to nothing about the document structure can be taken for granted. We have developed solutions where some of the heterogeneity issues are addressed. Our fragment selection algorithm selectively divides a heterogeneous document collection into equi-sized fragments with full-text content. If the content is considered too data-oriented, it is not accepted. The algorithm needs no information about element names. In addition, three techniques for fragment expansion are presented, all of which yield a 13--17% average improvement in average precision. These techniques and algorithms are among the first steps in developing document-type-independent indexing methods for the full text in heterogeneous XML collections.