Vector Of Locally Aggregated Descriptors Research Articles

A Hierarchical Utilization of Semantic Gradients and Scene Structure for Visual Place Recognition

Visual Place Recognition (VPR) is a fundamental element for long-term Simultaneous Localization and Mapping (SLAM) systems. For long-term VPR, severe appearance and viewpoint variations are inevitable. In this paper, we introduce a novel VPR system named Semantic Scene Structure Place Recognition (3SPR), inspired by the repeatability of semantic gradients and the scene structure of urban environments. Semantic gradients are densely sampled according to the sum of absolute gradients of all channels in the logits layer. Features of the semantic gradients in different layers are concatenated to exploit features’ characteristics at different levels. Based on partitions by vanishing points of road lines and Vector of Locally Aggregated Descriptors (VLAD), the Scene Structure VLAD (SSVLAD) is generated from concatenated features of the semantic gradients. Moreover, a local point group match method is used to enhance the spatial verification. Experimental results show that our method achieves state-of-the-art performance on the Oxford Robotcar dataset and the Synthia dataset.

SC_LPR: Semantically Consistent LiDAR Place Recognition Based on Chained Cascade Network in Long-Term Dynamic Environments.

In large-scale long-term dynamic environments, high-frequency dynamic objects inevitably lead to significant changes in the appearance of the scene at the same location at different times, which is catastrophic for place recognition (PR). Therefore, how to eliminate the influence of dynamic objects to achieve robust PR has universal practical value for mobile robots and autonomous vehicles. To this end, we suggest a novel semantically consistent LiDAR PR method based on chained cascade network, called SC_LPR, which mainly consists of a LiDAR semantic image inpainting network (LSI-Net) and a semantic pyramid Transformer-based PR network (SPT-Net). Specifically, LSI-Net is a coarse-to-fine generative adversarial network (GAN) with a gated convolutional autoencoder as the backbone. To effectively address the challenges posed by variable-scale dynamic object masks, we integrate the updated Transformer block with mask attention and gated trident block into LSI-Net. Sequentially, in order to generate a discriminative global descriptor representing the point cloud, we design an encoder with pyramid Transformer block to efficiently encode long-range dependencies and global contexts between different categories in the inpainted semantic image, followed by an augmented NetVALD, a generalized VLAD (Vector of Locally Aggregated Descriptors) layer that adaptively aggregates salient local features. Last but not least, we first attempt to create a LiDAR semantic inpainting dataset, called LSI-Dataset, to effectively validate the proposed method. Experimental comparisons show that our method not only improves semantic inpainting performance by about 6%, but also improves PR performance in dynamic environments by about 8% compared to the representative optimal baseline. LSI-Dataset will be publicly available at https://github.KD.LPR.com/.

VIEW GRAPH CONSTRUCTION FOR LARGE-SCALE UAV IMAGES: AN EVALUATION OF STATE-OF-THE-ART METHODS

Abstract. Structure from Motion (SfM) is a 3D reconstruction framework that has achieved great success on large-scale Unmanned Aerial Vehicle (UAV) images. Due to the high time consumption of feature matching, a matching candidate subset is obtained by image retrieval to improve efficiency. Bag of Word (BoW) based image retrieval has been widely used in SfM systems, but the large number of local features and the high dimension of the BoW vector cause the retrieval method time-consuming. Vector of Locally Aggregated Descriptors (VLAD) and learning-based NetVLAD perform well in image retrieval, and these vector representation methods are evaluated in this study. After images are transformed into vectors, Nearest Neighbour (NN) searching methods like Brute-force and KD-Tree are used to find similar images. But as the number of images and the vector dimension increase, Approximate Nearest Neighbour (ANN) searching methods like Hierarchical Navigable Small World (HNSW) and Locality-Sensitive Hashing (LSH) are considered to replace NN searching to avoid efficiency degradation. These vector searching methods are also evaluated in this study. The test results demonstrate that the optimal method VLAD with HNSW can speed up about 100 times in finding matching candidate subset. A view graph that guides scene partition and sub-scene reconstruction in parallel SfM can be created by the optimal method. With this view graph construction method, the efficiency of SfM is significantly improved.

Acoustic bird species classification under low SNR and small-scale dataset conditions

Bird species classification based on acoustic monitoring has attracted increasing attention. Two common methods, deep learning and feature extraction-based, are widely applied for bird sound classification. Although deep learning based methods have shown great success, they usually require an enormous amount of labeled data to train models, which could make these approaches prohibitive since collecting enough labeled samples of various species, especially rare birds, is a difficult task. On the other hand, feature extraction based methods are less dependent on the number of samples and have achieved remarkable classification performance for small-scale datasets, however they prove susceptible to signal-to-noise ratio (SNR). To address this problem, a new method is proposed for acoustic bird species classification in this work, which is well suited to low SNR and small-scale dataset conditions. Firstly, an adaptive threshold scheme based on the constant false alarm rate (CFAR) criterion is designed to more effectively detect potential bird sound segments in field recordings. Then, the local wavelet acoustic pattern (LWAP) and Mel-frequency cepstral coefficients (MFCC) features are extracted, which are further processed using improved vector of locally aggregated descriptors (VLAD) encoding. Finally, encoded LWAP and MFCC features are fused with dimensionality reduction and a classifier with weighted distance is used for classification. Corresponding feature space analysis upon feature fusion is also provided demonstrating explicit improvement in bird sounds discrimination under low SNR conditions. Experimental results on the dataset consisting of 11 bird species under low SNR and small-scale conditions reveal that the proposed method achieves considerable improvement in classification performance as compared with other feature extraction based methods. Meanwhile, our approach also outperforms state-of-the-art deep learning based methods.

Factor graph localization for mobile robots using Google Indoor Street View and CNN-based place recognition

This article proposes a mobile robot localization system developed using Google Indoor Street View (GISV) and Convolutional Neural Network (CNN)-based visual place recognition. The proposed localization system consists of two main modules. The first is a place recognition module based on GISV and a net Vector of Locally Aggregated Descriptors (VLAD)-based CNN. The second is a factor graph-based optimization module. In this work, we show that a CNN-based approach can be utilized to overcome the lack of visually distinct features in indoor environments and changes in images that can occur when using different cameras at different points in time for localization. The proposed CNN-based localization system is implemented using reference and query images obtained from two different sources (GISV and a camera attached to a mobile robot). It has been experimentally validated using a custom indoor dataset captured at the Memorial University of Newfoundland engineering building basement. The main results of this paper show that GISV-based place recognition reduces the percentage drift by 4% for the dataset and achieves a Root Mean Square Error (RMSE) of 2 m for position and 2.5° for orientation.

Data-Efficient Collaborative Decentralized Thermal-Inertial Odometry

We propose a system solution to achieve data-efficient, decentralized state estimation for a team of flying robots using thermal images and inertial measurements. Each robot can fly independently, and exchange data when possible to refine its state estimate. Our system front-end applies an online photometric calibration to refine the thermal images so as to enhance feature tracking and place recognition. Our system back-end uses a covariance-intersection fusion strategy to neglect the cross-correlation between agents so as to lower memory usage and computational cost. The communication pipeline uses Vector of Locally Aggregated Descriptors (VLAD) to construct a request-response policy that requires low bandwidth usage. We test our collaborative method on both synthetic and real-world data. Our results show that the proposed method improves by up to 46 % trajectory estimation with respect to an individual-agent approach, while reducing up to 89 % the communication exchange. Datasets and code are released to the public, extending the already-public JPL xVIO library.

Unifying Deep ConvNet and Semantic Edge Features for Loop Closure Detection

Loop closure detection is an important component of Simultaneous Localization and Mapping (SLAM). In this paper, a novel two-branch loop closure detection algorithm unifying deep Convolutional Neural Network (ConvNet) features and semantic edge features is proposed. In detail, we use one feature extraction module to extract both ConvNet and semantic edge features simultaneously. The deep ConvNet features are subjected to a Context Feature Enhancement (CFE) module in the global feature ranking branch to generate a representative global feature descriptor. Concurrently, to reduce the interference of dynamic features, the extracted semantic edge information of landmarks is encoded through the Vector of Locally Aggregated Descriptors (VLAD) framework in the semantic edge feature ranking branch to form semantic edge descriptors. Finally, semantic, visual, and geometric information is integrated by the similarity score fusion calculation. Extensive experiments on six public datasets show that the proposed approach can achieve competitive recall rates at 100% precision compared to other state-of-the-art methods.

Remote Sensing Image Retrieval Based on 3D-Local Ternary Pattern (LTP) Features and Non-subsampled Shearlet Transform (NSST) Domain Statistical Features

With the increasing popularity of high-resolution remote sensing images, the remote sensing image retrieval (RSIR) has always been a topic of major issue. A combined, global non-subsampled shearlet transform (NSST)-domain statistical features (NSSTds) and local three dimensional local ternary pattern (3D-LTP) features, is proposed for high-resolution remote sensing images. We model the NSST image coefficients of detail subbands using 2-state laplacian mixture (LM) distribution and its three parameters are estimated using Expectation-Maximization (EM) algorithm. We also calculate the statistical parameters such as subband kurtosis and skewness from detail subbands along with mean and standard deviation calculated from approximation subband, and concatenate all of them with the 2-state LM parameters to describe the global features of the image. The various properties of NSST such as multiscale, localization and flexible directional sensitivity make it a suitable choice to provide an effective approximation of an image. In order to extract the dense local features, a new 3D-LTP is proposed where dimension reduction is performed via selection of ‘uniform’ patterns. The 3D-LTP is calculated from spatial RGB planes of the input image. The proposed inter-channel 3D-LTP not only exploits the local texture information but the color information is captured too. Finally, a fused feature representation (NSSTds-3DLTP) is proposed using new global (NSSTds) and local (3D-LTP) features to enhance the discriminativeness of features. The retrieval performance of proposed NSSTds-3DLTP features are tested on three challenging remote sensing image datasets such as WHU-RS19, Aerial Image Dataset (AID) and PatternNet in terms of mean average precision (MAP), average normalized modified retrieval rank (ANMRR) and precision-recall (P-R) graph. The experimental results are encouraging and the NSSTds-3DLTP features leads to superior retrieval performance compared to many well known existing descriptors such as Gabor RGB, Granulometry, local binary pattern (LBP), Fisher vector (FV), vector of locally aggregated descriptors (VLAD) and median robust extended local binary pattern (MRELBP). For WHU-RS19 dataset, in terms of {MAP,ANMRR}, the NSSTds-3DLTP improves upon Gabor RGB, Granulometry, LBP, FV, VLAD and MRELBP descriptors by {41.93%,20.87%}, {92.30%,32.68%}, {86.14%,31.97%}, {18.18%,15.22%}, {8.96%,19.60%} and {15.60%,13.26%}, respectively. For AID, in terms of {MAP,ANMRR}, the NSSTds-3DLTP improves upon Gabor RGB, Granulometry, LBP, FV, VLAD and MRELBP descriptors by {152.60%,22.06%}, {226.65%,25.08%}, {185.03%,23.33%}, {80.06%,12.16%}, {50.58%,10.49%} and {62.34%,3.24%}, respectively. For PatternNet, the NSSTds-3DLTP respectively improves upon Gabor RGB, Granulometry, LBP, FV, VLAD and MRELBP descriptors by {32.79%, 10.34%}, {141.30%, 24.72%}, {17.47%,10.34%}, {83.20%,19.07%}, {21.56%,3.60%}, and {19.30%,0.48%} in terms of {MAP,ANMRR}. The moderate dimensionality of simple NSSTds-3DLTP allows the system to run in real-time.

A Novel Method of Aircraft Detection under Complex Background Based on Circular Intensity Filter and Rotation Invariant Feature.

Aircraft detection in remote sensing images (RSIs) has drawn widespread attention in recent years, which has been widely used in the military and civilian fields. While the complex background, variations of aircraft pose and size bring great difficulties to the effective detection. In this paper, we propose a novel aircraft target detection scheme based on small training samples. The scheme is coarse-to-fine, which consists of two main stages: region proposal and target identification. First, in the region proposal stage, a circular intensity filter, which is designed based on the characteristics of the aircraft target, can quickly locate the centers of multi-scale suspicious aircraft targets in the RSIs pyramid. Then the target regions can be extracted by adding bounding boxes. This step can get high-quality but few candidate regions. Second, in the stage of target identification, we proposed a novel rotation-invariant feature, which combines rotation-invariant histogram of oriented gradient and vector of locally aggregated descriptors (VLAD). The feature can characterize the aircraft target well by avoiding the impact of its rotation and can be effectively used to remove false alarms. Experiments are conducted on Remote Sensing Object Detection (RSOD) dataset to compare the proposed method with other advanced methods. The results show that the proposed method can quickly and accurately detect aircraft targets in RSIs and achieve a better performance.

Deep Convolutional Neural Network with KNN Regression for Automatic Image Annotation

Automatic image annotation is an active field of research in which a set of annotations are automatically assigned to images based on their content. In literature, some works opted for handcrafted features and manual approaches of linking concepts to images, whereas some others involved convolutional neural networks (CNNs) as black boxes to solve the problem without external interference. In this work, we introduce a hybrid approach that combines the advantages of both CNN and the conventional concept-to-image assignment approaches. J-image segmentation (JSEG) is firstly used to segment the image into a set of homogeneous regions, then a CNN is employed to produce a rich feature descriptor per area, and then, vector of locally aggregated descriptors (VLAD) is applied to the extracted features to generate compact and unified descriptors. Thereafter, the not too deep clustering (N2D clustering) algorithm is performed to define local manifolds constituting the feature space, and finally, the semantic relatedness is calculated for both image–concept and concept–concept using KNN regression to better grasp the meaning of concepts and how they relate. Through a comprehensive experimental evaluation, our method has indicated a superiority over a wide range of recent related works by yielding F1 scores of 58.89% and 80.24% with the datasets Corel 5k and MSRC v2, respectively. Additionally, it demonstrated a relatively high capacity of learning more concepts with higher accuracy, which results in N+ of 212 and 22 with the datasets Corel 5k and MSRC v2, respectively.

A Hierarchical Dual Model of Environment- and Place-Specific Utility for Visual Place Recognition

Visual Place Recognition (VPR) approaches have typically attempted to match places by identifying visual cues, image regions or landmarks that have high “utility” in identifying a specific place. But this concept of utility is not singular - rather it can take a range of forms. In this letter, we present a novel approach to deduce two key types of utility for VPR: the utility of visual cues `specific' to an environment, and to a particular place. We employ contrastive learning principles to estimate both the environment- and place-specific utility of Vector of Locally Aggregated Descriptors (VLAD) clusters in an unsupervised manner, which is then used to guide local feature matching through keypoint selection. By combining these two utility measures, our approach achieves state-of-the-art performance on three challenging benchmark datasets, while simultaneously reducing the required storage and compute time. We provide further analysis demonstrating that unsupervised cluster selection results in semantically meaningful results, that finer grained categorization often has higher utility for VPR than high level semantic categorization (e.g. building, road), and characterise how these two utility measures vary across different places and environments. Source code is made publicly available at https://github.com/Nik-V9/HEAPUtil.

Satellite and Scene Image Classification Based on Transfer Learning and Fine Tuning of ResNet50

Image classification has gained lot of attention due to its application in different computer vision tasks such as remote sensing, scene analysis, surveillance, object detection, and image retrieval. The primary goal of image classification is to assign the class labels to images according to the image contents. The applications of image classification and image analysis in remote sensing are important as they are used in various applied domains such as military and civil fields. Earlier approaches for remote sensing images and scene analysis are based on low-level feature representations such as color- and texture-based features. Vector of Locally Aggregated Descriptors (VLAD) and orderless Bag-of-Features (BoF) representations are the examples of mid-level approaches for remote sensing image classification. Recent trends for remote sensing and scene classification are focused on the use of Convolutional Neural Network (CNN). Keeping in view the success of CNN models, in this research, we aim to fine-tune ResNet50 by using network surgery and creation of network head along with the fine-tuning of hyperparameters. The learning of hyperparameters is tuned by using a linear decay learning rate scheduler known as piecewise scheduler. To tune the optimizer hyperparameter, Stochastic Gradient Descent with Momentum (SGDM) is used with the usage of weight learn and bias learn rate factor. Experiments and analysis are conducted on five different datasets, that is, UC Merced Land Use Dataset (UCM), RSSCN (the remote sensing scene classification image dataset), SIRI-WHU, Corel-1K, and Corel-1.5K. The analysis and competitive results exemplify that our proposed image classification-based model can classify the images in a more effective and efficient manner as compared to the state-of-the-art research.

Writer identification and writer retrieval based on NetVLAD with Re‐ranking

This paper addresses writer identification and writer retrieval which is considered as a challenging problem in the document analysis and recognition field. In this work, a novel pipeline is proposed for the problem at hand by employing a unified neural network architecture consisting of the ResNet-20 as a feature extractor and an integrated NetVLAD layer, inspired by the vector of locally aggregated descriptors (VLAD), in the head of the latter part. Having defined this architecture, the triplet semi-hard loss function is used to directly learn an embedding for individual input image patches. Subsequently, generalized max-pooling technique is employed for the aggregation of embedded descriptors of each handwritten image. Also, a novel re-ranking strategy is introduced for the task of identification and retrieval based on $k$-reciprocal nearest neighbors, and it is shown that the pipeline can benefit tremendously from this step. Experimental evaluation has been done on the three publicly available datasets: the ICDAR 2013, CVL, and KHATT datasets. Results indicate that while we perform comparably to the state-of-the-art on the KHATT, our writer identification and writer retrieval pipeline achieves superior performance on the ICDAR 2013 and CVL datasets in terms of mAP.

A Comprehensive Study on VLAD

Recently, the vector of locally aggregated descriptor (VLAD) has shown its great effectiveness in diverse computer vision tasks including image retrieval, Scene classification, and action recognition. Its great success stems from its powerful representation ability and computational efficiency. However, it remains unclear about its theoretical foundation and how it is connected to basic while important algorithms, e.g., the bag-of-words model and match kernels, and how its performance is affected by parameter configurations, e.g., normalization and pooling, which are also widely used in state-of-the-art algorithms based on local features. In this paper, with an aim to achieve the full capacity of VLAD, we conduct a comprehensive and in-depth study from both theoretical analysis and experimental practice perspectives. As a theoretical contribution, we provide a new formulation of VLAD via match kernels, which serves to connect VLAD with existing important encoding methods based on local features. As a contribution to the practical use of VLAD, we comprehensively investigate the roles and effects of the two widely-used operations in local feature encoding: normalization and pooling. To the best of our knowledge, our work provides the first comprehensive study on VLAD, which will not only enable a full understanding of it but also provide an important guidance for state-of-the-art algorithms based on local features. We have conducted extensive experiments on three benchmark datasets: Scene-15, Caltech 101 and PPMI for both image classification and action recognition.

Vector of Locally and Adaptively Aggregated Descriptors for Image Feature Representation

VLAD (Vector of Locally Aggregated Descriptors) has been widely adopted in image representation. However, the VLAD algorithm seeks for the algebraic sum of the residue vectors between the descriptors and the centroid of cluster they belong to, and this could decrease the discriminative power of feature representations. To this end, this paper originally proposes a VLAAD (Vector of Locally and Adaptively Aggregated Descriptors) framework to adaptively assign a weight to each residue vector. First, we compute the weights using the magnitude of each residue vector, and encapsulate the weighted VLAD block into ResNet to form an end-to-end Weighted NetVLAD method. To further enhance the discriminative power of the features, we subsequently replace the magnitude-based weight computation with a gating scheme to achieve automatic weight estimation. The enhanced version is named as Gated NetVLAD method. The experimental results on CIFAR-10, MNIST Digits, Pittsburgh Google street view and ImageNet-Dog datasets demonstrate the promotion in classification accuracy and retrieval mAP using VLAAD against several state-of-the-art methods.

KVGCN: A KNN Searching and VLAD Combined Graph Convolutional Network for Point Cloud Segmentation

Semantic segmentation of the sensed point cloud data plays a significant role in scene understanding and reconstruction, robot navigation, etc. This work presents a Graph Convolutional Network integrating K-Nearest Neighbor searching (KNN) and Vector of Locally Aggregated Descriptors (VLAD). KNN searching is utilized to construct the topological graph of each point and its neighbors. Then, we perform convolution on the edges of constructed graph to extract representative local features by multiple Multilayer Perceptions (MLPs). Afterwards, a trainable VLAD layer, NetVLAD, is embedded in the feature encoder to aggregate the local and global contextual features. The designed feature encoder is repeated for multiple times, and the extracted features are concatenated in a jump-connection style to strengthen the distinctiveness of features and thereby improve the segmentation. Experimental results on two datasets show that the proposed work settles the shortcoming of insufficient local feature extraction and promotes the accuracy (mIoU 60.9% and oAcc 87.4% for S3DIS) of semantic segmentation comparing to existing models.

Identification of Miao Embroidery in Southeast Guizhou Province of China Based on Convolution Neural Network

Abstract Miao embroidery of the southeast area of Guizhou province in China is a kind of precious intangible cultural heritage, as well as national costume handcrafts and textiles, with delicate patterns that require exquisite workmanship. There are various skills to make Miao embroidery; therefore, it is difficult to distinguish the categories of Miao embroidery if there is a lack of sufficient knowledge about it. Furthermore, the identification of Miao embroidery based on existing manual methods is relatively low and inefficient. Thus, in this work, a novel method is proposed to identify different categories of Miao embroidery by using deep convolutional neural networks (CNNs). Firstly, we established a Miao embroidery image database and manually assigned an accurate category label of Miao embroidery to each image. Then, a pre-trained deep CNN model is fine-tuned based on the established database to learning a more robust deep model to identify the types of Miao embroidery. To evaluate the performance of the proposed deep model for the application of Miao embroidery categories recognition, three traditional non-deep methods, that is, bag-of-words (BoW), Fisher vector (FV), and vector of locally aggregated descriptors (VLAD) are employed and compared in the experiment. The experimental results demonstrate that the proposed deep CNN model outperforms the compared three non-deep methods and achieved a recognition accuracy of 98.88%. To our best knowledge, this is the first one to apply CNNs on the application of Miao embroidery categories recognition. Moreover, the effectiveness of our proposed method illustrates that the CNN-based approach might be a promising strategy for the discrimination and identification of different other embroidery and national costume patterns.

FhVLAD: Fine-grained quantization and encoding high-order descriptor statistics for scalable image retrieval

We are interested in the encoding of local descriptors of an image (e.g. SIFT) to design a compact representation vector and thereby address scalable image retrieval. We revisit the implicit design choices in the popular vector of locally aggregated descriptors (VLAD), which aggregates the residuals of descriptors to the codewords. VLAD’s use of a coarse codebook and first-order descriptor statistics in residual computation results in less discriminative residuals. To address this problem, we propose a division of codebook feature space using a novel fine-grained quantization strategy. After quantization, we embed the resulting residuals with high-order statistics of descriptor distribution. Experiments on three challenging image retrieval datasets (INRIA Holidays, UKBench, Oxford 5k) confirm the improved discriminative power of our novel encoding method called FhVLAD. We observe superior accuracy to baseline and competitive performance to state-of-the-art techniques with a limited increase in dimension.

A transfer learning approach for improved classification of carbon nanomaterials from TEM images.

The extensive use of carbon nanomaterials such as carbon nanotubes/nanofibers (CNTs/CNFs) in industrial settings has raised concerns over the potential health risks associated with occupational exposure to these materials. These exposures are commonly in the form of CNT/CNF-containing aerosols, resulting in a need for a reliable structure classification protocol to perform meaningful exposure assessments. However, airborne carbonaceous nanomaterials are very likely to form mixtures of individual nano-sized particles and micron-sized agglomerates with complex structures and irregular shapes, making structure identification and classification extremely difficult. While manual classification from transmission electron microscopy (TEM) images is widely used, it is time-consuming due to the lack of automation tools for structure identification. In the present study, we applied a convolutional neural network (CNN) based machine learning and computer vision method to recognize and classify airborne CNT/CNF particles from TEM images. We introduced a transfer learning approach to represent images by hypercolumn vectors, which were clustered via K-means and processed into a Vector of Locally Aggregated Descriptors (VLAD) representation to train a softmax classifier with the gradient boosting algorithm. This method achieved 90.9% accuracy on the classification of a 4-class dataset and 84.5% accuracy on a more complex 8-class dataset. The developed model established a framework to automatically detect and classify complex carbon nanostructures with potential applications that extend to the automated structural classification for other nanomaterials.

Aggregated Deep Global Feature Representation for Breast Cancer Histopathology Image Classification

Convolutional neural networks (CNNs), successfully used in a great number of medical image analysis applications, have also achieved the state-of-the-art performance in breast cancer histopathology image (BCHI) classification problem recently. However, due to the large varieties among within-class images and insufficient data volume, it is still a challenge to obtain more competitive results by using deep CNN models alone. In this paper, we aim to explore the combination of CNN models with a milestone feature representation method in visual tasks, i.e., vector of locally aggregated descriptors (VLAD), for the BCHI classification, and further propose a novel aggregated deep global feature representation (ADGFR) for this problem. ADGFR adopts the deep features that are extracted from the fully connected layer to form an individual descriptor vector, and augments input images to generate different descriptors for achieving the final aggregated descriptor vector. The individual descriptor vector can effectively keep the global features of benign and malignant image, whose discriminability is further reinforced by the aggregate operation, leading to the more discriminant capability of ADGFR for BCHI. Extensive experiments on the public Break His dataset illuminate that our ADGFR can achieve the optimal classification accuracies of 95.05% at image level and 95.50% at patient level, respectively.