Fine Classes Research Articles

Improving spatiotemporal image fusion incorporating unmixing step by considering the point spread function effect

ABSTRACT In this paper, an improving spatiotemporal image fusion (STIF) incorporating unmixing step by considering the point spread function (PSF) effect model (CPSF) is proposed to address the problem that unmixing accuracy in STIF models incorporating unmixing step is vulnerable to the PSF effect. First, an unsupervised clustering algorithm is used to cluster the fine image at previous date into a land cover map, and the coarse class fraction matrix is generated from this land cover map. Second, the fine class fraction matrix uncontaminated by the PSF effect is calculated by upsampling the coarse class fraction matrix using the Area-to-Point Kriging (ATPK) method under the PSF constraint. Third, this fine class fraction matrix is convolved with the ideal square wave filter to downsample, and then normalized to obtain the coarse class fraction matrix without the influence of the PSF effect. Finally, the CPSF-processed class fraction matrix is used to replace the original one in the following unmixing and prediction process to generate the fine resolution image at the prediction date. Experiments on two remote sensing data sets show that the proposed method can effectively improve the accuracy of spatiotemporal image fusion methods incorporating unmixing step.

CRSANet: Class Representations Self-Attention network for the segmentation of thyroid nodules

Segmenting accurate thyroid nodules in medical ultrasound images is always non-trivial due to the large variation in size, shape and texture of nodules, and also the presence of surrounding tissues and organs with similar intensity values. Existing algorithms solve this by designing multi-scale spatial context modeling modules and introducing the emerging self-attention based paradigm, while failing to learn the class-wise relational context around nodules. To this end, we propose a novel coarse-to-fine framework, namely, CRSANet, which leverages both class-wise relational context and spatial multi-scale context for comprehensive context perception. Specifically, we first introduce the concept of Class Representations, which describe the overall representation of thyroid nodules in ultrasound images from a categorical perspective. Secondly, we propose a backbone-free dual-branch Class Representation Self-Attention Module (CRSA) to refine the coarse segmentation results from segmentation backbones. Concretely, Fine Class Region (FCR) branch is proposed to impose consistency between pixels and their corresponding Class Representations to correct the wrongly classified pixels. Meanwhile, Class Contextual Representation (CCR) branch is introduced to enhance the ability of pixel representations with the corresponding Class Representations. We evaluate the effectiveness and generality of CRSANet using three types of backbone networks. In the experiments, our proposed U-CRSANet achieves a mIoU of 78.86% and 81.35%, an accuracy of 97.11% and 98.63%, a Dice value of 74.26% and 87.57%, respectively, which outperforms most state-of-the-art methods on the public DDTI and TN-SCUI datasets.

Image classification for historical documents: a study on Chinese local gazetteers

Abstract We present a novel approach for automatically classifying illustrations from historical Chinese local gazetteers using modern deep learning techniques. Our goal is to facilitate the digital organization and study of a large quantity of digitized local gazetteers. We evaluate the performance of eight state-of-the-art deep neural networks on a dataset of 4,309 manually labeled and organized images of Chinese local gazetteer illustrations, grouped into three coarse categories and nine fine classes according to their contents. Our experiments show that DaViT achieved the highest classification accuracy of 93.9 per cent and F1-score of 90.6 per cent. Our results demonstrate the effectiveness of deep learning models in accurately recognizing and categorizing historical local gazetteer illustrations. We also developed a user-friendly web service to enable researchers easy access to the developed models. The potential for extending this method to other collections of scanned documents beyond Chinese local gazetteers makes a significant contribution to the study of visual materials in the arts and history in the digital humanities field. The dataset used in this study is publicly available and can be used for further research in the field.

Features of Gold Placer Formation in the Ukrainian Crystalline Shield And Southern Ukraine

The presence of numerous native gold deposits of the Ukrainian Crystalline Shield (UCS), as well as the spread of placer gold manifestations, determines the high probability of the formation here of gold placers with industrial parameters. The main factors and geological conditions that contribute to the formation of gold placers are manifested at the UCS. An important reason for the failures of previous geological works is the outdated scientific and methodological platform, which involved the search for typical gold-bearing deposits with metal of gravity size classes. They did not take into account the possibility of significant movement of placer gold in the geologically clay water flow and the formation of long-distance placers. The traditional Schlich method used in the study did not make it possible to determine the actual gold content when particles with a size smaller than 0.25 mm predominate in the sample, which is precisely the particle size of the primary sources of the UCS. The applied means of research could not be effective in the absence of the necessary analytical base for determining the gold content, modern methods of studying the geological structure and tectonics of the territories. The negative results did not provide any real grounds for further specialized large-scale research and prospecting for placer gold. Today, it is getting obvious that placer gold of small and fine classes can easily move over considerable distances in the water flow and form accumulations not only in boulder-pebble material, but also in pelitic sediments. Due to repeated block movements, erosion and redepositing of gold-bearing accumulations in secondary reservoirs could occur. In this aspect, neotectonic studies (using materials from space surveys) become the main predictive and prospecting method for identifying promising areas of placer gold concentration in the region. When carrying out fieldwork, drilling and modern methods of laboratory analysis of gold samples should be used. The results of many years of research give hope for the discovery of real placer gold deposits in the Northern Black Sea region in the near future.

Multi-Stage Semantic Segmentation Quantifies Fragmentation of Small Habitats at a Landscape Scale

Land cover (LC) maps are used extensively for nature conservation and landscape planning, but low spatial resolution and coarse LC schemas typically limit their applicability to large, broadly defined habitats. In order to target smaller and more-specific habitats, LC maps must be developed at high resolution and fine class detail using automated methods that can efficiently scale to large areas of interest. In this work, we present a Machine Learning approach that addresses this challenge. First, we developed a multi-stage semantic segmentation approach that uses Convolutional Neural Networks (CNNs) to classify LC across the Peak District National Park (PDNP, 1439 km2) in the UK using a detailed, hierarchical LC schema. High-level classes were predicted with 95% accuracy and were subsequently used as masks to predict low-level classes with 72% to 92% accuracy. Next, we used these predictions to analyse the degree and distribution of fragmentation of one specific habitat—wet grassland and rush pasture—at the landscape scale in the PDNP. We found that fragmentation varied across areas designated as primary habitat, highlighting the importance of high-resolution LC maps provided by CNN-powered analysis for nature conservation.

MACHINE LEARNING IN THE DIAGNOSIS OF FOLLICULAR LYMPHOMA

Objective To implement a machine-learning (ML) model to assist pathologists in the differentiation of follicular hyperplasia (FH) and follicular lymphoma (FL). Study Design Whole slide images from 10 patients with FH and 10 patients with FL were manually annotated and fragmented into 21,585 patches (FH=13,399 and FL=8,186) of 299 × 299 pixels. The convolutional neural network (CNN) VGGNet was re-trained using Python 3.6 and other open-source libraries for machine learning and image processing (TensorFlow, Keras, Scikit-Learn, and OpenCV). The training and validation were carried out for 10 epochs until accuracy stabilized and validation loss reduced its variation. Results The total processing time for the CNN training was 753s. Different metrics could be obtained through the confusion matrix, emphasizing a high training accuracy of 98% and F1-score of 82%. Sensitivity and specificity were 74.8% and 91.8%, respectively. The receiver operating characteristic curve of 94% showed the fine class separation ability of the CNN. Conclusion The ML model used in this study is feasible to differentiate FH and FL. Additional CNN training and validation in bigger/multicenter datasets may generate AI-assisted tools to aid FL diagnosis. To implement a machine-learning (ML) model to assist pathologists in the differentiation of follicular hyperplasia (FH) and follicular lymphoma (FL). Whole slide images from 10 patients with FH and 10 patients with FL were manually annotated and fragmented into 21,585 patches (FH=13,399 and FL=8,186) of 299 × 299 pixels. The convolutional neural network (CNN) VGGNet was re-trained using Python 3.6 and other open-source libraries for machine learning and image processing (TensorFlow, Keras, Scikit-Learn, and OpenCV). The training and validation were carried out for 10 epochs until accuracy stabilized and validation loss reduced its variation. The total processing time for the CNN training was 753s. Different metrics could be obtained through the confusion matrix, emphasizing a high training accuracy of 98% and F1-score of 82%. Sensitivity and specificity were 74.8% and 91.8%, respectively. The receiver operating characteristic curve of 94% showed the fine class separation ability of the CNN. The ML model used in this study is feasible to differentiate FH and FL. Additional CNN training and validation in bigger/multicenter datasets may generate AI-assisted tools to aid FL diagnosis.

Automatic detection of prostate cancer grades and chronic prostatitis in biparametric MRI

Background and objective:With emerging evidence to improve prostate cancer (PCa) screening, multiparametric magnetic prostate imaging is becoming an essential noninvasive component of the diagnostic routine. Computer-aided diagnostic (CAD) tools powered by deep learning can help radiologists interpret multiple volumetric images. In this work, our objective was to examine promising methods recently proposed in the multigrade prostate cancer detection task and to suggest practical considerations regarding model training in this context.Methods:We collected 1647 fine-grained biopsy-confirmed findings, including Gleason scores and prostatitis, to form a training dataset. In our experimental framework for lesion detection, all models utilized 3D nnU-Net architecture that accounts for anisotropy in the MRI data. First, we explore an optimal range of b-values for diffusion-weighted imaging (DWI) modality and its effect on the detection of clinically significant prostate cancer (csPCa) and prostatitis using deep learning, as the optimal range is not yet clearly defined in this domain. Next, we propose a simulated multimodal shift as a data augmentation technique to compensate for the multimodal shift present in the data. Third, we study the effect of incorporating the prostatitis class alongside cancer-related findings at three different granularities of the prostate cancer class (coarse, medium, and fine) and its impact on the detection rate of the target csPCa. Furthermore, ordinal and one-hot encoded (OHE) output formulations were tested.Results: An optimal model configuration with fine class granularity (prostatitis included) and OHE has scored the lesion-wise partial Free-Response Receiver Operating Characteristic (FROC) area under the curve (AUC) of 1.94 (CI 95%: 1.76–2.11) and patient-wise ROC AUC of 0.874 (CI 95%: 0.793–0.938) in the detection of csPCa. Inclusion of the auxiliary prostatitis class has demonstrated a stable relative improvement in specificity at a false positive rate (FPR) of 1.0 per patient, with an increase of 3%, 7%, and 4% for coarse, medium, and fine class granularities.Conclusions: This paper examines several configurations for model training in the biparametric MRI setup and proposes optimal value ranges. It also shows that the fine-grained class configuration, including prostatitis, is beneficial for detecting csPCa. The ability to detect prostatitis in all low-risk cancer lesions suggests the potential to improve the quality of the early diagnosis of prostate diseases. It also implies an improved interpretability of the results by the radiologist.

Fine sediment dynamics over a gravel bar. Part 1: Validation of a new image-based segmentation method

In this paper, we present a new image processing algorithm to analyse fine sediment deposits on gravel bars. The method is based on pictures taken from the river bank with an average resolution of a few centimetres per pixel. The proposed image processing algorithm consists of two steps. In the first step, the raw image is orthorectified with a 5 cm pixel size. In second step, we used an original combination of two well-known image techniques to classify pixel. First, a high-pass filter is applied in the Fourier domain to correct grey-value potential bias, i.e. to merge wet and dry deposits. Then, an unsupervised classification process based on K-means clustering is conducted to partition image pixels into K classes. The area of fine sediment deposits is calculated by summing the area of each class multiplied by weighting parameters (surface area percentage of fine deposits for each class). The value of K and the weighting parameters are calibrated using local close range photos of the bar surface representing the spatial variation of surface fine sediment distribution. The algorithm is then validated with field measurement using the finest class (d<4 mm) of Wolman pebble count method. The impact of image resolution is investigated: the algorithm was found suitable for pixel size between 1 and 10 cm. The robustness of the present algorithm to images taken at different times and with different illuminations is studied by comparing it to two methods from the literature. The new method yields the most robust results with an average standard deviation lower than 5%. It yields larger surface of fine sediment since it includes all possible patches around coarse particles. Such results imply that the algorithm can be applied on a long image series to help us understand the temporal dynamics of fine sediment deposits on the bar surface.

Application of spiral separation technology in coal sludge processing

The purpose of the study is to analyze the current state of the problem concerning the concentration technology of wash house coal sludge, comparison of the results of sludge dump coal processing on various types of processing equipment as well as formulation of recommendations for spiral separation method application for coal sludge enrichment. The main research methods include generalization and analysis of scientific and technical information, study of the results of technological tests on spiral separation. The problem of sludge concentration is relevant due to the increase in the content of fine classes in mined coal and, consequently, in coal sludge. In real practice most of the sludge is fuel for energy. Having analyzed the experimental part of the research, the authors established the efficiency of coal and rock fraction separation using the technology of coal sludge spiral separation. The need for preliminary classification of the source material before spiral separation is determined. It is found out that spiral sludge separators have a number of technological advantages as compared to spiral mineral separators, due to a wider range of enriched material fineness. The analysis of the performed tests made it possible to establish the application prospects of the spiral separation technology in coal sludge processing. The recommendations are given on the use of spiral separators manufactured by the Research and Production Company Spirit, LLC for the enrichment of sludge products including tailings, dump sludge of the work site of the Wash House of the Branch Open Pit Cheremkhovugol of Vostsibugol Company, LLC.

Development and testing of a combined process based on the use of a flat-bottomed hydrocyclone and a multi-frequency screen

The article provides a brief overview of the development of technology and equipment for the preparation of finely disseminated ores of ferrous and non-ferrous metals for further processing. Fine screening processes and equipment for crushed ores are considered, including in combination with centrifugal hydraulic classification. The purpose of the work was to improve hydraulic screening performance and efficiency for crushed ores. In order to optimize the preparation of finely disseminated ores for processing, a combined preliminary hydraulic classification process was developed for crushed ores based on the use of a flat-bottomed hydrocyclone and a multi-frequency screen. A pilot plant was used to test the process. The article contains qualitative and quantitative performance indicators for the hydraulic screening of crushed chromium ore with a particle size of 300&ndash;0 &mu;m. The total recovery of all size classes below 300 &mu;m into the undersize of the multi-frequency screen was 86.42 % and the recovery of finer ore classes (below 150 &mu;m) was 91.09 to 92.97 %. The resulting combined process using a flat-bottomed hydrocyclone and a multi-frequency screen improves the total equipment performance, ensures high recoveries of fine classes into the undersize, extends the service life of fine sieves and reduces water consumption for oversize irrigation. This combined process is also recommended for use in hydraulic screening of magnetite concentrates at mining and processing plants in order to improve the iron content, as well as in closed fine grinding circuits in order to prevent overgrinding of valuable minerals and reduce the circulating load on the screens.

Synthetic aircraft RS image modelling based on improved conditional GAN joint embedding network

Developing an efficient and quality remote sensing (RS) technology using volume and efficient modelling in different aircraft RS images is challenging. Generative models serve as a natural and convenient simulation method. Because aircraft types belong to the fine class under the rough class, the issue of feature entanglement may occur while modelling multiple aircraft classes. Our solution to this issue was a novel first-generation realistic aircraft type simulation system (ATSS-1) based on the RS images. It realised fine modelling of the seven aircraft types based on a real scene by establishing an adaptive weighted conditional attention generative adversarial network and joint geospatial embedding (GE) network. An adaptive weighted conditional batch normalisation attention block solved the subclass entanglement by reassigning the intra-class-wise characteristic responses. Subsequently, an asymmetric residual self-attention module was developed by establishing a remote region asymmetric relationship for mining the finer potential spatial representation. The mapping relationship between the input RS scene and the potential space of the generated samples was explored through the GE network construction that used the selected prior distribution z, as an intermediate representation. A public RS dataset (OPT-Aircraft_V1.0) and two public datasets (MNIST and Fashion-MNIST) were used for simulation model testing. The results demonstrated the effectiveness of ATSS-1, promoting further development of realistic automatic RS simulation.

Apatite-staffelite ore flotation efficiency improvement&#x0D; using two-stage slurry thickening

The studies were performed suggesting that the&#x0D; cause of P2O5 losses during apatite-staffelite ores (ASO)&#x0D; treatment are due to non-selective flocculation of fine&#x0D; classes during flotation. When using strong flocculants,&#x0D; special preparation of condensed slurries is necessary,&#x0D; ensuring their deflocculation before the flotation process.&#x0D; A scheme and mode of preparation of fine classes for the&#x0D; flotation process have been developed, including thickening&#x0D; of the classification overflows using strong anionic&#x0D; flocculants and deflocculation of the thickened product&#x0D; before the flotation process with reagents-dispersants&#x0D; used in the basic flotation mode. A mode of preparation&#x0D; of slimes of ASO ores for flotation is proposed, including&#x0D; thickening of the discharge of the classification&#x0D; operation using the anionic flocculant “Praestol-2540”,&#x0D; conditioning of the condensed product with additions of&#x0D; liquid glass and caustic soda in a ratio of 1 : 1, dilution&#x0D; and re-thickening of deflocculated slimes, consolidation&#x0D; and flotation thickened sludge and sand. The big laboratory&#x0D; tests have shown that the application of the developed&#x0D; regime provides a total increase in the extraction&#x0D; of P2O5 from ore from 70,1 to 71,5 % with an increase&#x0D; in the P2O5 content in apatite concentrate from 37,1 to&#x0D; 37,8 %, which makes the developed technology promising&#x0D; for processing refractory ASO at Kovdorsky GOK.

Fine-grain beta diversity in Palaearctic open vegetation: variability within and between biomes and vegetation types

Aims: To quantify how fine-grain (within-plot) beta diversity differs among biomes and vegetation types. Study area: Palaearctic biogeographic realm. Methods: We extracted 4,654 nested-plot series with at least four different grain sizes between 0.0001 m² and 1,024 m² from the GrassPlot database spanning broad geographic and ecological gradients. Next, we calculated the slope parameter (z-value) of the power-law species–area relationship (SAR) to use as a measure of multiplicative beta diversity. We did this separately for vascular plants, bryophytes and lichens and for the three groups combined (complete vegetation). We then tested whether z-values differed between biomes, ecological-physiognomic vegetation types at coarse and fine levels and phytosociological classes. Results: We found that z-values varied significantly among biomes and vegetation types. The explanatory power of area for species richness was highest for vascular plants, followed by complete vegetation, bryophytes and lichens. Within each species group, the explained variance increased with typological resolution. In vascular plants, adjusted R2 was 0.14 for biomes, but reached 0.50 for phytosociological classes. Among the biomes, mean z-values were particularly high in the Subtropics with winter rain (Mediterranean biome) and the Dry tropics and subtropics. Natural grasslands had higher z-values than secondary grasslands. Alpine and Mediterranean vegetation types had particularly high z-values whereas managed grasslands with benign soil and climate conditions and saline communities were characterised by particularly low z-values. Conclusions: In this study relating fine-grain beta diversity to typological units, we found distinct patterns. As we explain in a conceptual figure, these can be related to ultimate drivers, such as productivity, stress and disturbance, which can influence z-values via multiple pathways. The provided means, medians and quantiles of z-values for a wide range of typological entities provide benchmarks for local to continental studies, while calling for additional data from under-represented units. Syntaxonomic references: Mucina et al. (2016) for classes occurring in Europe; Ermakov (2012) for classes restricted to Asia. Abbreviations: ANOVA = analysis of variance; EDGG = Eurasian Dry Grassland Group; SAR = species-area relationship.

Hysteretic Implications for Graded Bed Load Sediment Transport in Symmetrical Hydrograph Flows

Differential parametric values associated with bed load sediment transport, that result at the same discharge levels on the rising and falling limbs of a flood hydrograph, are usually defined as bed load hysteresis. This hysteresis in bed load sediment transport rates is of considerable interest in the field of fluvial hydraulics. Within this study, a series of well-defined, symmetrical hydrograph flows are generated over a graded, mobile sediment bed to fully examine the hysteresis of the resulting bed load sediment transport in terms of the threshold of motion, and differential bed load transport rates and bed load yields during the hydrographs. The experiments are conducted in a titling flume without sediment supply specified at the upstream inlet, thereby representing typical river reach conditions immediately downstream of a dam that are exclusively subject to net in-channel bed degradation from sediment transport initiated during flood events. Our results show that the fractional bed load transport of defined fine, medium and coarse size classes within the graded sediment bed generally display clockwise, no/mixed and counter-clockwise hysteresis patterns, respectively, with clockwise hysteresis most commonly found for the coarse size class mobilised by hydrographs with long durations. By contrast, counter-clockwise hysteresis is usually observed for fine size class transported by hydrographs with short durations. Accordingly, the corresponding reference stresses for each size class vary between different hydrographs and are primarily controlled by the hydrograph flashiness (i.e. unsteadiness) and magnitude (i.e. total water work). Moreover, it is shown that the hysteresis effect, particularly for those size classes and hydrograph combinations that result in clockwise and counter-clockwise behaviour, should be fully accounted for when reproducing bed load transport rates using separate-limb based method. Finally, we investigate the relative fractions of the overall bed load yields generated during the rising and falling limbs of all symmetrical hydrographs (i.e. the bed load yield ratio), which are found to be primarily dependent on bed load transport hysteresis. Finally, the relationship between the bed load yield ratio and the ratio of reference stresses for the fractional sediment motion of each size class on both limbs is found to follow a power law.

ФЛОТАЦИОННЫЕ РЕАГЕНТЫ ДЛЯ ИЗВЛЕЧЕНИЯ ТОНКОВКРАПЛЕННОГО ЗОЛОТА ИЗ ТРУДНООБОГАТИМЫХ РУД И ТЕХНОГЕННЫХ ПРОДУКТОВ

At the present stage of development of the mineral resource base of precious metals, the wider involvement of refractory ores and technogenic products in the mineral processing is playing a very important role. It is due to these types of mineral resources the main Russian and foreign companies plan to get the increase in the production of gold and other precious metals. Refractory ores are characterized by low content and uneven distribution of valuable components, emulsion dissemination of gold, complex material composition, flotation active silicate and carbonaceous rock components. The main technologies for their processing include combined schemes that provide intensive preliminary preparation of ore, fine and ultrafine grinding. Creation of effective reagents for flotation of fine classes of noble metals, selective depression of gold-free sulfides, like pyrite, pyrrhotite, arsenopyrite, and rock minerals is one of the recognized innovative solutions for intensifying the enrichment process of this type of raw material. The studies performed at IPKON RAS in the framework of the scientific school of Academician V.A. Chanturiya helped to develop and substantiate the novel selective reagents-collectors of dithiocarbamate type (DEDTCm, OPDTC, DBDTC, MDTC) and herbal modifiers that possess complexing properties in relation to noble metals and provide an increase in the recovery of valuable components of gold by 5-7%. A set of analytical methods - scanning electron (LEO 1420VP, Carl Zeiss) and laser microscopy (KEYNCE VK-9700, Shimadzu), X-ray fluorescence spectroscopy (ARL ADVANT’X), ICP –MS (Agilent 725 ICP-OES, Agilent Technologies), X-ray diffractometry (DRON-2.0, XRD 7000, Shimadzu), potentiometry, titrimetric analysis, methods of mathematical statistics helped to study the adsorption layer of the reagents on natural minerals and model sulfides with micro- and nanosized gold. Using an original technique for quantitative assessment of the reagent coating by image analysis (SLM), the surface area of the mineral occupied by the reagent was calculated. A method for selective flotation of gold-containing sulfides using OPDTC and oak bark extract has been developed and tested in the separation of the collective gold-arsenic concentrate of the Olympiada deposit. It showed the possibility of obtaining a gold-pyrite concentrate with an arsenic content of less than 2%. The prospect of using plant modifiers for gold recovery during flotation of complex ores has been substantiated.

Characterization of Hybrid Materials Prepared by Sol-Gel Method for Biomedical Implementations. A Critical Review.

The interaction between tissues and biomaterials (BM) has the purpose of improving and replacing anatomical parts of the human body, avoiding the occurrence of adverse reactions in the host organism. Unfortunately, the early failure of implants cannot be currently avoided, since neither a good mixture of mechanical and chemical characteristics of materials nor their biocompatibility has been yet achieved. Bioactive glasses are recognized to be a fine class of bioactive substances for good repair and replacement. BM interact with living bones through the formation of a hydroxyapatite surface layer that is analogous to bones. Bioglasses’ composition noticeably affects their biological properties, as does the synthesis method, with the best one being the versatile sol-gel technique, which includes the change of scheme from a ‘sol’ fluid into a ‘gel’. This process is widely used to prepare many materials for biomedical implants (e.g., hip and knee prostheses, heart valves, and ceramic, glassy and hybrid materials to serve as carriers for drug release). Nanoparticles prepared by the sol-gel method are interesting systems for biomedical implementations, and particularly useful for cancer therapy. This review provides many examples concerning the synthesis and characterization of the above-mentioned materials either taken from literature and from recently prepared zirconia/polyethylene glycol (PEG) hybrids, and the corresponding results are extensively discussed.

End-to-End 6DoF Pose Estimation From Monocular RGB Images

We present a conceptually simple framework for 6DoF object pose estimation, especially for autonomous driving scenarios. Our approach can efficiently detect the traffic participants from a monocular RGB image while simultaneously regressing their 3D translation and rotation vectors. The proposed method 6D-VNet, extends the Mask R-CNN by adding customised heads for predicting vehicle's finer class, rotation and translation. It is trained end-to-end compared to previous methods. Furthermore, we show that the inclusion of translational regression in the joint losses is crucial for the 6DoF pose estimation task, where object translation distance along longitudinal axis varies significantly, e.g., in autonomous driving scenarios. Additionally, we incorporate the mutual information between traffic participants via a modified non-local block to capture the spatial dependencies among the detected objects. As opposed to the original non-local block implementation, the proposed weighting modification takes the spatial neighbouring information into consideration whilst counteracting the effect of extreme gradient values. We evaluate our method on the challenging real-world Pascal3D+ dataset and our 6D-VNet reaches the 1st place in ApolloScape challenge 3D Car Instance task (Apolloscape, 2018), (Huang et al., 2018).

Extraction and Analysis of Finer Impervious Surface Classes in Urban Area

Impervious surfaces (IS), the most common land cover in urban areas, not only provide convenience to the city, but also exert significant negative environmental impacts, thereby affecting the ecological environment carrying capacity of urban agglomerations. Most of the current research considers IS as a single land-cover type, yet this does not fully reflect the complex physical characteristics of various IS types. Therefore, limited information for urban micro-ecology and urban fine management can be provided through one IS land-cover type. This study proposed a finer IS classification scheme and mapped the detailed IS fraction in Guangzhou City, China using Landsat imagery. The IS type was divided into seven finer classes, including blue steel, cement, asphalt, other impervious surface, and other metal, brick, and plastic. Classification results demonstrate that finer IS can be well extracted from the Landsat imagery as all root mean square errors (RMSE) are less than 15%. Specially, the accuracies of asphalt, plastic, and cement are better than other finer IS types with the RMSEs of 7.99%, 8.48%, and 9.92%, respectively. Quantitative analyses illustrate that asphalt, other impervious surface, and brick are the dominant IS types in the study area with the percentages of 9.68%, 6.27%, and 4.45%, respectively, and they are mainly located in Yuexiu, Liwan, Haizhu, and Panyu districts. These results are valuable for research into urban fine management and can support the detailed analysis of urban micro-ecology.

Use of computer technologies in the study of architectural decoration in fine arts classes

Use of computer technologies in the study of architectural decoration in fine arts classes

The problem of student compositional vision development in fine arts classes

The problem of student compositional vision development in fine arts classes