Transfer learning-based Fashion Image Classification using Hybrid 2D-CNN and ImageNet Neural Network

Many algorithms have been developed as a result of recent advances in machine learning to handle a variety of challenges. In recent years, the most popular transfer learning method has allowed researchers and engineers to run experiments with minimal computing and time resources. To tackle the challenges of classification, product identification, product suggestion, and picture-based search, this research proposed a transfer learning strategy for Fashion image classification based on hybrid 2D-CNN pretrained by VGG-16 and AlexNet. Pre-processing, feature extraction, and classification are the three parts of the proposed system's implementation. We used the Fashion MNIST dataset, which consists of 50,000 fashion photos that have been classified. Training and validation datasets have been separated. In comparison to other conventional methodologies, the suggested transfer learning approach has higher training and validation accuracy and reduced loss. Keywords: Machine Learning, Transfer Learning, Convolutional Neural Network, Image Classification, VGG16, AlexNet, 2D CNN.

RESEARCH ON QUALITY CONTROL METHOD OF LAND COVER CLASSIFICATION DATA ORIENTED TO NATIONAL GEOGRAPHIC CONDITION MONITORING

Aiming at the quality problems of the land cover classification data, such as missing update of changed spots, overrun of spot collection accuracy, incorrect spot classification, and inconsistencies between the filling of the change types of map spots and technical regulations, had been found during the quality inspection of the national geographic condition monitoring in China. This paper analyzed the key production factors, processes and causes that affect the process quality of the land cover classification data, and proposed the concept of the whole production quality control, and discussed the quality control measures to improve the land cover classification product. The research based on the new requirements of the whole-process quality control system in order to solve the quality problem-oriented production quality control method, could improve the quality of land cover classification product and provided reference for the quality control of the production process of related engineering projects in the future.

Budget constrained interactive search for multiple targets

Interactive graph search leverages human intelligence to categorize target labels in a hierarchy, which is useful for image classification, product categorization, and database search. However, many existing interactive graph search studies aim at identifying a single target optimally, and suffer from the limitations of asking too many questions and not being able to handle multiple targets. To address these two limitations, in this paper, we study a new problem of <u>b</u>udget constrained <u>i</u>nteractive <u>g</u>raph <u>s</u>earch for <u>m</u>ultiple targets called kBM-IGS problem. Specifically, given a set of multiple targets T in a hierarchy and two parameters k and b , the goal is to identify a k -sized set of selections S , such that the closeness between selections S and targets T is as small as possible, by asking at most a budget of b questions. We theoretically analyze the updating rules and design a penalty function to capture the closeness between selections and targets. To tackle the kBM-IGS problem, we develop a novel framework to ask questions using the best vertex with the largest expected gain, which provides a balanced trade-off between target probability and benefit gain. Based on the kBM-IGS framework, we first propose an efficient algorithm STBIS to handle the SingleTarget problem, which is a special case of kBM-IGS. Then, we propose a dynamic programming based method kBM-DP to tackle the MultipleTargets problem. To further improve efficiency, we propose two heuristic but efficient algorithms, kBM-Topk and kBM-DP+. Experiments on large real-world datasets with ground-truths verify both the effectiveness and efficiency of our algorithms.

Approaches to the Formation of Intangible Resources Based on the Innovations Introduction in the Higher Education System

The article is aimed at determining approaches to the identification and formation of evaluation indicators as to the production, availability and use of intangible resources, including intellectual ones, as a result of innovative changes in the activities of higher education institutions and their reform on an innovative basis. The role of introduction of innovations in the higher education system for improving the innovative capacity and competitiveness of enterprises and the national economy as a whole is defined. Based on the analysis of statistical data, the need to reform domestic higher education institutions on an innovative basis is identified. The role of implementation of the key performance indicators (KPI) system for identification and evaluation of intangible assets of both intellectual and information nature as a result of innovative changes is substantiated. A general approach to assessing the results of innovations on the basis of giving a tuple look to the model for computing KPI indicators is proposed. The analysis of strategies of innovative development of leading higher education institutions of Ukraine for the implementation of innovative changes is carried out. The identified innovations are grouped into the main types according to the Oslo classification: product-related, technological, marketing and organizational. Each type of innovation is considered in the plane of educational, scientific and international activities. The main directions of innovative changes for each of these groups are allocated and technologies for innovative changes are determined. It is noted how the changes in question affect the formation of the intellectual resource of higher education institutions and business environment entities.

Detailed Lacustrine Calving Iceberg Inventory from Very High Resolution Optical Imagery and Object-Based Image Analysis

In the field of iceberg and glacier calving studies, it is important to collect comprehensive datasets of populations of icebergs. Particularly, calving of lake-terminating glaciers has been understudied. The aim of this work is to present an object-based method of iceberg detection and to create an inventory of icebergs located in a proglacial lagoon of San Quintín glacier, Northern Patagonia Icefield, Chile. This dataset is created using high-resolution WorldView-2 imagery and a derived DEM. We use it to briefly discuss the iceberg size distribution and area–volume scaling. Segmentation of the multispectral imagery produced a map of objects, which were classified with use of Random Forest supervised classification algorithm. An intermediate classification product was corrected with a ruleset exploiting contextual information and a watershed algorithm that was used to divide multiple touching icebergs into separate objects. Common theoretical heavy-tail statistical distributions were tested as descriptors of iceberg area and volume distributions. Power law models were proposed for the area–volume relationship. The proposed method performed well over the open lake detecting correctly icebergs in all size bands except 5–15 m2 where many icebergs were missed. A section of the lagoon with ice melange was not reliably mapped due to uniformity of the area in the imagery and DEM. The precision of the DEM limited the scaling effort to icebergs taller than 1.7 m and larger than 99 m2, despite the inventory containing icebergs as small as 4 m2. The work demonstrates viability of object-based analysis for lacustrine iceberg detection and shows that the statistical properties of iceberg population at San Quintín glacier match those of populations produced by tidewater glaciers.

Distributions and radiative forcings of various cloud types based on active and passive satellite datasets – Part 1: Geographical distributions and overlap of cloud types

Based on four year' 2B-CLDCLASS-Lidar (Radar-Lidar) cloud classification product from CloudSat, we analyze the geographical distributions of different cloud types and their co-occurrence frequency across different seasons, moreover, utilize the vertical distributions of cloud type to further evaluate the cloud overlap assumptions. The statistical results show that more high clouds, altocumulus, stratocumulus or stratus and cumulus are identified in the Radar-Lidar cloud classification product compared to previous results from Radar-only cloud classification (2B-CLDCLASS product from CloudSat). In particularly, high clouds and cumulus cloud fractions increased by factors 2.5 and 4–7, respectively. The new results are in more reasonable agreement with other datasets (typically the International Satellite Cloud Climatology Project (ISCCP) and surface observer reports). Among the cloud types, altostratus and altocumulus are more popular over the arid/semi-arid land areas of the Northern and Southern Hemispheres, respectively. These features weren't observed by using the ISCCP D1 dataset. For co-occurrence of cloud types, high cloud, altostratus, altocumulus and cumulus are much more likely to co-exist with other cloud types. However, stratus/stratocumulus, nimbostratus and convective clouds are much more likely to exhibit individual features. After considering the co-occurrence of cloud types, the cloud fraction based on the random overlap assumption is underestimated over the vast ocean except in the west-central Pacific Ocean warm pool. Obvious overestimations are mainly occurring over land areas in the tropics and subtropics. The investigation therefore indicates that incorporate co-occurrence information of cloud types based on Radar-Lidar cloud classification into the overlap assumption schemes used in the current GCMs possible be able to provide an better predictions for vertically projected total cloud fraction.

Effective UV surface albedo of seasonally snow-covered lands

At ultraviolet wavelengths the albedo of most natural surfaces is small with the striking exception of snow and ice. Therefore, snow cover is a major challenge for various applications based on radiative transfer modelling. The aim of this work was to determine the characteristic effective UV range surface albedo of various land cover types when covered by snow. First we selected 1 by 1 degree sample regions that met three criteria: the sample region contained dominantly subpixels of only one land cover type according to the 8 km global land cover classification product from the University of Maryland; the average slope of the sample region was less than 2 degrees according to the USGS's HYDRO1K slope data; the sample region had snow cover in March according to the NSIDC Northern Hemisphere weekly snow cover data. Next we generated 1 by 1 degree gridded 360 nm surface albedo data from the Nimbus-7 TOMS Lambertian equivalent reflectivity data, and used them to construct characteristic effective surface albedo distributions for each land cover type. The resulting distributions showed that each land cover type experiences a characteristic range of surface albedo values when covered by snow. The result is explained by the vegetation that extends upward beyond the snow cover and masks the bright snow covered surface.

Effective UV surface albedo of seasonally snow-covered lands

At ultraviolet wavelengths the albedo of most natural surfaces is small with the striking exception of snow and ice. Therefore, snow cover is a major challenge for various applications based on radiative transfer modelling. The aim of this work was to determine the characteristic effective UV range surface albedo of various land cover types when covered by snow. First we selected 1 by 1 degree sample regions that met three criteria: the sample region contained dominantly subpixels of only one land cover type according to the 8 km global land cover classification product from the University of Maryland; the average slope of the sample region was less than 2 degrees according to the USGS's HYDRO1K slope data; the sample region had snow cover in March according to the NSIDC Northern Hemisphere weekly snow cover data. Next we generated 1 by 1 degree gridded 360 nm surface albedo data from the Nimbus-7 TOMS Lambertian equivalent reflectivity data, and used them to construct characteristic effective surface albedo distributions for each land cover type. The resulting distributions showed that each land cover type experiences a characteristic range of surface albedo values when covered by snow. The result is explained by the vegetation that extends upward beyond the snow cover and masks the bright snow covered surface.