Search

Recently, there has been an upsurge in publicly available remote sensing image classification datasets. Standard CNNs and pre-trained architectures have been applied for scene classification tasks. However, transfer learning models accept specific image dimensions as the minimum required size for their respective image input layers. Depending on the size of the input image, the final feature map might not contain the discriminative information needed for accurately classifying the dataset categories. The proposed technique effectively enables and enhances a transfer learning model (Xception) to be applied to scene classification tasks. The model works on an adaptive framework that interpolates images and selects an appropriate dilation layer to enhance the quality of extracted features for improved classification. This approach is evaluated on the EuroSAT, a dataset with images of 64x64 pixels, UCM and AID datasets, respectively. We recorded 98.55%, 99.22%, and 96.15% accuracy for the EuroSAT, UCM, and AID datasets, respectively. Our model and the reported results have opened the potential of the Xception, which in our view, has not been given its fair share of attention, despite its efficient parameter utilisation. [ABSTRACT FROM AUTHOR]

In order to capture the high-level concepts in high spatial resolution remote sensing (HSR) imagery, scene classification based on a latent Dirichlet allocation (LDA) model, a generative topic model, is a practical method to bridge the semantic gaps between the low-level features and the high-level concepts of HSR imagery. In the previous work, LDA has been considered as a scene classifier, namely C-LDA, and multiple LDA models for each scene class are built separately, where the scene class is determined by a maximum likelihood rule. The C-LDA strategy disregards the correlations between the generative topic spaces of the different scene classes. In this paper, two novel strategies of scene classification based on LDA are proposed to consider the correlations between the generative topic spaces of the different scene classes by sharing the topic spaces for all the scene classes. One of the proposed strategies utilizes LDA as part of the classifier, namely P-LDA, which generates the topic space from all the training images. A discriminative classifier (e.g., support vector machine, SVM) is also employed as the other classification part of P-LDA. The other proposed strategy employs LDA as the topic feature extractor, namely F-LDA, which generates the topic space from all the training and test images, and utilizes a discriminative classifier to classify the topic features. The experimental results using aerial orthophotographs show that the performances of the two proposed strategies for scene classification based on LDA are better than the traditional C-LDA method. [ABSTRACT FROM PUBLISHER]

Scene classification is essential in Content Based Image Retrieval system which processes large amount of data. But this retrieval system failed to produce good classification result due to less relevant visual features used for classification task. In this work, we aim to categorize scenes into indoor versus outdoor using relevant low level features such as color and texture which help to improve the classification performance. The proposed method uses statistical feature computed from HSV color model as color feature, DCT coefficients as texture feature and entropy computed with UV. A simple non-parametric K-nearest neighbor classifier is used in conjunction with this low level features to categorize scene into indoor versus outdoor. Since these image features exhibit a distinctive disparity between images containing indoor or outdoor scenes the proposed method achieves better performance in terms of classification accuracy about 81.5% when less number training images are used. The proposed method is evaluated on IITM-SCID 2 (scene classification image database) as well as 2011 images collected from the web. [ABSTRACT FROM PUBLISHER]

In this paper we espouse to build a scoring procedure that evaluates the similarity between a sentence and an image. This approach is attractive, because it is symmetric: given an image (resp. sentence), one can search for the best sentence (resp. image) in a large set. Furthermore, because the learning cost is independent from the amount of data, the system has scalability, which makes it useful with large-scale data. [ABSTRACT FROM PUBLISHER]

In this paper we present a methodology to use Gabor response features for real-time visual road environment classification. Processing Gabor filters using hardware solely dedicated to this task enables improved real-time texture classification. Using such hardware enables us to successfully extract Gabor feature information for a four-class road environment classification problem. We used summary histogram as an intermediate level of texture representation prior to final classification. Overall we obtain a maximally correct classification circa 98%, outperforming prior work in the field. [ABSTRACT FROM PUBLISHER]

Scene images classification is one of the most important methods in robot navigation in unstructured environment. In this paper, a scene images classification method based on visual feature fusion is proposed. First, the scene image is divided into fixed-size blocks. Second, computes the DCT coefficients of each blocks and runs a wavelet-like decomposition, means and variances are extracted in each component of the decomposed sub-blanks. Third, color moments are extracted and fused with DCT texture-based features to describe each block. Finally, a Gaussian Mixture Model (GMM) is employed for training and classification. [ABSTRACT FROM AUTHOR]

In classification problem, single classifier may not fully catch the dataset's information. Thus, an ensemble method based on Support Vector Machine (SVM) is proposed in this paper for image scene classification. First, Scale Invariant Feature Transform (SIFT) is used to extract the features of the images, and the SIFT features are clustered to form a visual vocabulary. Then, the SIFT features of each image are compared with this visual vocabulary to calculate the appearance frequencies of the visual words, which consist of the Bag-of-Words (BOW) model descriptions of the image. Probabilistic Latent Semantic Analysis (PLSA) is used to exploit the latent semantic features on the basis of the BOW model, and SVM classifier is then trained by these latent semantic features. These processes repeat N times and N different SVM classifiers are trained. Finally, they are used to classify the testing images, and the ensemble of the N different classification results are calculated as the final result. Experiments show that our method can be effectively applied to the scene classification problem, and the accuracy could be improved with a certain degree of robustness. [ABSTRACT FROM PUBLISHER]

Natural scene recognition and classification have received considerable attention in the computer vision community due to its challenging nature. Significant intra-class variations have largely limited the accuracy of scene categorization tasks: a holistic representation forces matching in strict spatial confinement; whereas a bag of features representation ignores the order or spatial layout of the scene completely, resulting in a loss of scene logic. In this paper, we present a novel method, called ARP (Angular Radial Partitioning) Gist, to classify the scene. Experiments show that the proposed method has improved recognition accuracy by better representing the structure in a scene and striking a balance between spatial confinement and freedom. [ABSTRACT FROM PUBLISHER]

Due to the importance of feature extraction and scene representation in classification tasks, this paper presents an approach for unsupervised feature learning using Independent Subspace Analysis. The optimization process of feature bases is incorporated into the framework of incremental learning to cope with the learning difficulty with large or dynamic samples. The proposed method could automatically learn image features and accomplish scene classification with Spatial Pyramid Matching model. Also, the influence of related parameters in optimization and classification is discussed. Experiment shows the proposed method constructs efficient scene description and outperforms several previous methods in classification on OT scene dataset. [ABSTRACT FROM PUBLISHER]

In this paper, we introduce a new dataset for scene classification based on camera metadata. We classify the most common scenes that have been researched much recently. This dataset consists of 12 scene categories. Each category contains 500 to 2000 images. Most images are high resolution such as 2000×2000 pixels. The images in the dataset are original, namely, each image brings with a camera metadata (EXIF). Various types, metadata cues of photos, strict definitions among scenes are characteristic factors that make this dataset a very challenging testbed for photo classification. We supply the scene photos together with scene labeling, as well as the EXIF information extraction via methodology, and we apply the dataset into sementic scene classification up to now. [ABSTRACT FROM PUBLISHER]

The use of local features for image representation has been proven very effective for a variety of visual tasks such as object localization and scene classification. However, local image features carry little semantic information which is potentially not enough for high level visual tasks. To solve this problem, in this paper, we propose to use a supervised semantic image representation for scene classification, where an image is represented as a response histogram. This response histogram is a combination of the prediction of pre-trained generic object classifiers and classifiers generated by supervised learning. Besides, the use of sparsity constraints makes the proposed representation more efficient and effective to compute. Performances on the UIUC-Sports dataset, the MIT Indoor scene dataset and the Scene-15 dataset demonstrate the effectiveness of the proposed method. [ABSTRACT FROM PUBLISHER]

The hierarchical structures that exist in natural scenes have been utilized for many tasks in computer vision. The basic idea is that instead of using strictly low level features it is possible to combine them into higher level hierarchical structures. These higher level structures provide a more specific feature and can thus lead to better results in classification or detection. Although most previous work has focused on hierarchical combinations of low level features, hierarchical structures exist on higher levels as well. In this work we attempt to automatically discover these higher level structures by finding meaningful object groups using the Minimum Description Length (MDL) principle. We then use these structures for scene classification and show that we can achieve a higher accuracy rate using them. [ABSTRACT FROM PUBLISHER]

This paper introduces several novel color, shape and texture-based image descriptors for scene image classification with applications to image search and retrieval. Specifically, first, a new 3-Dimensional Local Binary Pattern (3DLBP) descriptor is proposed for color image local feature extraction. Second, a new shape descriptor (HaarHOG) is introduced by combining Haar wavelet transformation and Histogram of Oriented Gradients (HOG). Third, these descriptors are fused using an optimal feature representation technique to generate a robust 3-Dimensional LBP-HaarHOG (3DLH) descriptor that can perform well on different scene image categories. Finally, the Enhanced Fisher Model (EFM) is applied for discriminatory feature extraction and the nearest neighbor classification rule is used for image classification. The proposed descriptors and fusion technique are evaluated using three grand challenge datasets: the MIT Scene dataset, the UIUC Sports Event dataset, and a part of the Caltech 256 dataset. [ABSTRACT FROM PUBLISHER]

This paper proposes a hybrid algorithm based on improved LLE and adaptive k-means for visual codebook generation in tourism scene classification. Firstly, we construct the improved LLE algorithm to get lower dimensional and compressed image feature representations. Then we form the adaptive k-means clustering algorithm to generate the visual codebook. Finally, we use the visual codebook histogram to represent the samples and train the SVM classifier for scene classification task. Experiments are conducted on a Beijing tourism scene dataset to evaluate the performance of the hybrid algorithm. Experimental results show that our algorithm can effectively improve the robustness of the visual codebook and result in a satisfying performance of scene classification. [ABSTRACT FROM PUBLISHER]