Search

Mongolia is among the countries undergoing rapid urbanization, and its temporary nomadic dwellings-known as Ger-have expanded into urban areas. Ger settlements in cities are increasingly recognized as slums by their socio-economic deprivation. The distinctive circular, tent-like shape of gers enables their detection through very-high-resolution satellite imagery. We develop a computer vision algorithm to detect gers in Ulaanbaatar, the capital of Mongolia, utilizing satellite images collected from 2015 to 2023. Results reveal that ger settlements have been displaced towards the capital's peripheral areas. The predicted slum ratio based on our results exhibits a significant correlation (r = 0.84) with the World Bank's district-level poverty data. Our nationwide extrapolation suggests that slums may continue to take up one-fifth of the population after the COVID-19 pandemic, contrary to other official predictions that anticipated a decline. We discuss the potential of machine learning on satellite imagery in providing insights into urbanization patterns and monitoring the Sustainable Development Goals., Comment: 20 pages

Mechanisms are designed to perform functions in various fields. Often, there is no unique mechanism that performs a well-defined function. For example, vehicle suspensions are designed to improve driving performance and ride comfort, but different types are available depending on the environment. This variability in design makes performance comparison difficult. Additionally, the traditional design process is multi-step, gradually reducing the number of design candidates while performing costly analyses to meet target performance. Recently, AI models have been used to reduce the computational cost of FEA. However, there are limitations in data availability and different analysis environments, especially when transitioning from low-fidelity to high-fidelity analysis. In this paper, we propose a multi-fidelity design framework aimed at recommending optimal types and designs of mechanical mechanisms. As an application, vehicle suspension systems were selected, and several types were defined. For each type, mechanism parameters were generated and converted into 3D CAD models, followed by low-fidelity rigid body dynamic analysis under driving conditions. To effectively build a deep learning-based multi-fidelity surrogate model, the results of the low-fidelity analysis were analyzed using DBSCAN and sampled at 5% for high-cost flexible body dynamic analysis. After training the multi-fidelity model, a multi-objective optimization problem was formulated for the performance metrics of each suspension type. Finally, we recommend the optimal type and design based on the input to optimize ride comfort-related performance metrics. To validate the proposed methodology, we extracted basic design rules of Pareto solutions using data mining techniques. We also verified the effectiveness and applicability by comparing the results with those obtained from a conventional deep learning-based design process.

Molecules have a number of distinct properties whose importance and application vary. Often, in reality, labels for some properties are hard to achieve despite their practical importance. A common solution to such data scarcity is to use models of good generalization with transfer learning. This involves domain experts for designing source and target tasks whose features are shared. However, this approach has limitations: i). Difficulty in accurate design of source-target task pairs due to the large number of tasks, and ii). corresponding computational burden verifying many trials and errors of transfer learning design, thereby iii). constraining the potential of foundation modeling of multi-task molecular property prediction. We address the limitations of the manual design of transfer learning via data-driven bi-level optimization. The proposed method enables scalable multi-task transfer learning for molecular property prediction by automatically obtaining the optimal transfer ratios. Empirically, the proposed method improved the prediction performance of 40 molecular properties and accelerated training convergence.

Training deep learning models on limited data while maintaining generalization is one of the fundamental challenges in molecular property prediction. One effective solution is transferring knowledge extracted from abundant datasets to those with scarce data. Recently, a novel algorithm called Geometrically Aligned Transfer Encoder (GATE) has been introduced, which uses soft parameter sharing by aligning the geometrical shapes of task-specific latent spaces. However, GATE faces limitations in scaling to multiple tasks due to computational costs. In this study, we propose a task addition approach for GATE to improve performance on target tasks with limited data while minimizing computational complexity. It is achieved through supervised multi-task pre-training on a large dataset, followed by the addition and training of task-specific modules for each target task. Our experiments demonstrate the superior performance of the task addition strategy for GATE over conventional multi-task methods, with comparable computational costs., Comment: 11 pages, 5 figures, Accepted at AI for Science Workshop at 41st International Conference on Machine Learning

The fifth Sloan Digital Sky Survey (SDSS-V) Local Volume Mapper (LVM) is a wide-field integral field unit (IFU) survey that uses an array of four 160 mm fixed telescopes with siderostats to minimize the number of moving parts. Individual telescope observes the science field or calibration field independently and is synchronized with the science exposure. We developed the LVM Acquisition and Guiding Package (LVMAGP) optimized telescope control software program for LVM observations, which can simultaneously control four focusers, three K-mirrors, one fiber selector, four mounts (siderostats), and seven guide cameras. This software is built on a hierarchical architecture and the SDSS framework and provides three key sequences: autofocus, field acquisition, and autoguide. We designed and fabricated a proto-model siderostat to test the telescope pointing model and LVMAGP software. The mirrors of the proto-model were designed as an isogrid open-back type, which reduced the weight by 46% and enabled reaching thermal equilibrium quickly. Additionally, deflection due to bolting torque, self-gravity, and thermal deformation was simulated, and the maximum scatter of the pointing model induced by the tilt of optomechanics was predicted to be $4'.4$, which can be compensated for by the field acquisition sequence. We performed a real sky test of LVMAGP with the proto-model siderostat and obtained field acquisition and autoguide accuracies of $0''.38$ and $1''.5$, respectively. It met all requirements except for the autoguide specification, which will be resolved by more precise alignment among the hardware components at Las Campanas Observatory., Comment: 29 pages, 15 figures, in press at Journal of Astronomical Telescopes, Instruments, and Systems

Weakly-Supervised Group Activity Recognition (WSGAR) aims to understand the activity performed together by a group of individuals with the video-level label and without actor-level labels. We propose Flow-Assisted Motion Learning Network (Flaming-Net) for WSGAR, which consists of the motion-aware actor encoder to extract actor features and the two-pathways relation module to infer the interaction among actors and their activity. Flaming-Net leverages an additional optical flow modality in the training stage to enhance its motion awareness when finding locally active actors. The first pathway of the relation module, the actor-centric path, initially captures the temporal dynamics of individual actors and then constructs inter-actor relationships. In parallel, the group-centric path starts by building spatial connections between actors within the same timeframe and then captures simultaneous spatio-temporal dynamics among them. We demonstrate that Flaming-Net achieves new state-of-the-art WSGAR results on two benchmarks, including a 2.8%p higher MPCA score on the NBA dataset. Importantly, we use the optical flow modality only for training and not for inference.

Molecular datasets often suffer from a lack of data. It is well-known that gathering data is difficult due to the complexity of experimentation or simulation involved. Here, we leverage mutual information across different tasks in molecular data to address this issue. We extend an algorithm that utilizes the geometric characteristics of the encoding space, known as the Geometrically Aligned Transfer Encoder (GATE), to a multi-task setup. Thus, we connect multiple molecular tasks by aligning the curved coordinates onto locally flat coordinates, ensuring the flow of information from source tasks to support performance on target data., Comment: 7 pages, 3 figures, 2 tables

Panoramic Activity Recognition (PAR) seeks to identify diverse human activities across different scales, from individual actions to social group and global activities in crowded panoramic scenes. PAR presents two major challenges: 1) recognizing the nuanced interactions among numerous individuals and 2) understanding multi-granular human activities. To address these, we propose Social Proximity-aware Dual-Path Network (SPDP-Net) based on two key design principles. First, while previous works often focus on spatial distance among individuals within an image, we argue to consider the spatio-temporal proximity. It is crucial for individual relation encoding to correctly understand social dynamics. Secondly, deviating from existing hierarchical approaches (individual-to-social-to-global activity), we introduce a dual-path architecture for multi-granular activity recognition. This architecture comprises individual-to-global and individual-to-social paths, mutually reinforcing each other's task with global-local context through multiple layers. Through extensive experiments, we validate the effectiveness of the spatio-temporal proximity among individuals and the dual-path architecture in PAR. Furthermore, SPDP-Net achieves new state-of-the-art performance with 46.5\% of overall F1 score on JRDB-PAR dataset.

Mechanisms are essential components designed to perform specific tasks in various mechanical systems. However, designing a mechanism that satisfies certain kinematic or quasi-static requirements is a challenging task. The kinematic requirements may include the workspace of a mechanism, while the quasi-static requirements of a mechanism may include its torque transmission, which refers to the ability of the mechanism to transfer power and torque effectively. In this paper, we propose a deep learning-based generative model for generating multiple crank-rocker four-bar linkage mechanisms that satisfy both the kinematic and quasi-static requirements aforementioned. The proposed model is based on a conditional generative adversarial network (cGAN) with modifications for mechanism synthesis, which is trained to learn the relationship between the requirements of a mechanism with respect to linkage lengths. The results demonstrate that the proposed model successfully generates multiple distinct mechanisms that satisfy specific kinematic and quasi-static requirements. To evaluate the novelty of our approach, we provide a comparison of the samples synthesized by the proposed cGAN, traditional cVAE and NSGA-II. Our approach has several advantages over traditional design methods. It enables designers to efficiently generate multiple diverse and feasible design candidates while exploring a large design space. Also, the proposed model considers both the kinematic and quasi-static requirements, which can lead to more efficient and effective mechanisms for real-world use, making it a promising tool for linkage mechanism design.

This paper proposes a framework that optimizes the linkage mechanism of the quasi-serial manipulator for target tasks. This process is explained through a case study of 2-degree-of-freedom linkage mechanisms, which significantly affect the workspace of the quasi-serial manipulator. First, a vast quasi-serial mechanism is generated with a workspace satisfying a target task and it converts it into a 3D CAD model. Then, the workspace and required torque performance of each mechanism are evaluated through kinematic and dynamic analysis. A deep learning-based surrogate model is leveraged to efficiently predict mechanisms and performance during the optimization process. After model training, a multi-objective optimization problem is formulated under the mechanical and dynamic conditions of the manipulator. The design goal of the manipulator is to recommend quasi-serial mechanisms with optimized kinematic (workspace) and dynamic (joint torque) performance that satisfies the target task. To investigate the underlying physics from the obtained Pareto solutions, various data mining techniques are performed to extract design rules that can provide practical design guidance. Finally, the manipulator was designed in detail for realization with 3D printed parts, including topology optimization. Also, the task-based optimized manipulator is verified through a payload test. Based on these results, the proposed framework has the potential for other real applications as realized cases and provides a reasonable design plan through the design rule extraction.

Due to the distinctive characteristics of sensors, each modality exhibits unique physical properties. For this reason, in the context of multi-modal action recognition, it is important to consider not only the overall action content but also the complementary nature of different modalities. In this paper, we propose a novel network, named Modality Mixer (M-Mixer) network, which effectively leverages and incorporates the complementary information across modalities with the temporal context of actions for action recognition. A key component of our proposed M-Mixer is the Multi-modal Contextualization Unit (MCU), a simple yet effective recurrent unit. Our MCU is responsible for temporally encoding a sequence of one modality (e.g., RGB) with action content features of other modalities (e.g., depth and infrared modalities). This process encourages M-Mixer network to exploit global action content and also to supplement complementary information of other modalities. Furthermore, to extract appropriate complementary information regarding to the given modality settings, we introduce a new module, named Complementary Feature Extraction Module (CFEM). CFEM incorporates sepearte learnable query embeddings for each modality, which guide CFEM to extract complementary information and global action content from the other modalities. As a result, our proposed method outperforms state-of-the-art methods on NTU RGB+D 60, NTU RGB+D 120, and NW-UCLA datasets. Moreover, through comprehensive ablation studies, we further validate the effectiveness of our proposed method., Comment: arXiv admin note: substantial text overlap with arXiv:2208.11314

Transfer learning is a crucial technique for handling a small amount of data that is potentially related to other abundant data. However, most of the existing methods are focused on classification tasks using images and language datasets. Therefore, in order to expand the transfer learning scheme to regression tasks, we propose a novel transfer technique based on differential geometry, namely the Geometrically Aligned Transfer Encoder (GATE). In this method, we interpret the latent vectors from the model to exist on a Riemannian curved manifold. We find a proper diffeomorphism between pairs of tasks to ensure that every arbitrary point maps to a locally flat coordinate in the overlapping region, allowing the transfer of knowledge from the source to the target data. This also serves as an effective regularizer for the model to behave in extrapolation regions. In this article, we demonstrate that GATE outperforms conventional methods and exhibits stable behavior in both the latent space and extrapolation regions for various molecular graph datasets., Comment: 12+11 pages, 6+1 figures, 0+7 tables

The ongoing modernization of the power system, involving new equipment installations and upgrades, exposes the power system to the introduction of malware into its operation through supply chain attacks. Supply chain attacks present a significant threat to power systems, allowing cybercriminals to bypass network defenses and execute deliberate attacks at the physical layer. Given the exponential advancements in machine intelligence, cybercriminals will leverage this technology to create sophisticated and adaptable attacks that can be incorporated into supply chain attacks. We demonstrate the use of reinforcement learning for developing intelligent attacks incorporated into supply chain attacks against generation control devices. We simulate potential disturbances impacting frequency and voltage regulation. The presented method can provide valuable guidance for defending against supply chain attacks., Comment: 7 pages, conference, IEEE International Conference on Machine Learning and Applications (ICMLA) 2023

Deep learning has made significant strides in video understanding tasks, but the computation required to classify lengthy and massive videos using clip-level video classifiers remains impractical and prohibitively expensive. To address this issue, we propose Audio-Visual Glance Network (AVGN), which leverages the commonly available audio and visual modalities to efficiently process the spatio-temporally important parts of a video. AVGN firstly divides the video into snippets of image-audio clip pair and employs lightweight unimodal encoders to extract global visual features and audio features. To identify the important temporal segments, we use an Audio-Visual Temporal Saliency Transformer (AV-TeST) that estimates the saliency scores of each frame. To further increase efficiency in the spatial dimension, AVGN processes only the important patches instead of the whole images. We use an Audio-Enhanced Spatial Patch Attention (AESPA) module to produce a set of enhanced coarse visual features, which are fed to a policy network that produces the coordinates of the important patches. This approach enables us to focus only on the most important spatio-temporally parts of the video, leading to more efficient video recognition. Moreover, we incorporate various training techniques and multi-modal feature fusion to enhance the robustness and effectiveness of our AVGN. By combining these strategies, our AVGN sets new state-of-the-art performance in multiple video recognition benchmarks while achieving faster processing speed., Comment: ICCV 2023

We introduce Sketch-based Video Object Localization (SVOL), a new task aimed at localizing spatio-temporal object boxes in video queried by the input sketch. We first outline the challenges in the SVOL task and build the Sketch-Video Attention Network (SVANet) with the following design principles: (i) to consider temporal information of video and bridge the domain gap between sketch and video; (ii) to accurately identify and localize multiple objects simultaneously; (iii) to handle various styles of sketches; (iv) to be classification-free. In particular, SVANet is equipped with a Cross-modal Transformer that models the interaction between learnable object tokens, query sketch, and video through attention operations, and learns upon a per-frame set matching strategy that enables frame-wise prediction while utilizing global video context. We evaluate SVANet on a newly curated SVOL dataset. By design, SVANet successfully learns the mapping between the query sketches and video objects, achieving state-of-the-art results on the SVOL benchmark. We further confirm the effectiveness of SVANet via extensive ablation studies and visualizations. Lastly, we demonstrate its transfer capability on unseen datasets and novel categories, suggesting its high scalability in real-world applications

Standard multi-modal models assume the use of the same modalities in training and inference stages. However, in practice, the environment in which multi-modal models operate may not satisfy such assumption. As such, their performances degrade drastically if any modality is missing in the inference stage. We ask: how can we train a model that is robust to missing modalities? This paper seeks a set of good practices for multi-modal action recognition, with a particular interest in circumstances where some modalities are not available at an inference time. First, we study how to effectively regularize the model during training (e.g., data augmentation). Second, we investigate on fusion methods for robustness to missing modalities: we find that transformer-based fusion shows better robustness for missing modality than summation or concatenation. Third, we propose a simple modular network, ActionMAE, which learns missing modality predictive coding by randomly dropping modality features and tries to reconstruct them with the remaining modality features. Coupling these good practices, we build a model that is not only effective in multi-modal action recognition but also robust to modality missing. Our model achieves the state-of-the-arts on multiple benchmarks and maintains competitive performances even in missing modality scenarios. Codes are available at https://github.com/sangminwoo/ActionMAE., Comment: AAAI 2023 (Oral); Code: https://github.com/sangminwoo/ActionMAE

In drug discovery, aqueous solubility is an important pharmacokinetic property which affects absorption and assay availability of drug. Thus, in silico prediction of solubility has been studied for its utility in virtual screening and lead optimization. Recently, machine learning (ML) methods using experimental data has been popular because physics-based methods like quantum mechanics and molecular dynamics are not suitable for high-throughput tasks due to its computational costs. However, ML method can exhibit over-fitting problem in a data-deficient condition, and this is the case for most chemical property datasets. In addition, ML methods are regarded as a black box function in that it is difficult to interpret contribution of hidden features to outputs, hindering analysis and modification of structure-activity relationship. To deal with mentioned issues, we developed Bayesian graph neural networks (GNNs) with the self-attention readout layer. Unlike most GNNs using self-attention in node updates, self-attention applied at readout layer enabled a model to improve prediction performance as well as to identify atom-wise importance, which can help lead optimization as exemplified for three FDA-approved drugs. Also, Bayesian inference enables us to separate more or less accurate results according to uncertainty in solubility prediction task We expect that our accurate, reliable and interpretable model can be used for more careful decision-making and various applications in the development of drugs.

In multi-modal action recognition, it is important to consider not only the complementary nature of different modalities but also global action content. In this paper, we propose a novel network, named Modality Mixer (M-Mixer) network, to leverage complementary information across modalities and temporal context of an action for multi-modal action recognition. We also introduce a simple yet effective recurrent unit, called Multi-modal Contextualization Unit (MCU), which is a core component of M-Mixer. Our MCU temporally encodes a sequence of one modality (e.g., RGB) with action content features of other modalities (e.g., depth, IR). This process encourages M-Mixer to exploit global action content and also to supplement complementary information of other modalities. As a result, our proposed method outperforms state-of-the-art methods on NTU RGB+D 60, NTU RGB+D 120, and NW-UCLA datasets. Moreover, we demonstrate the effectiveness of M-Mixer by conducting comprehensive ablation studies., Comment: IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2023

Natural Language Video Grounding (NLVG) aims to localize time segments in an untrimmed video according to sentence queries. In this work, we present a new paradigm named Explore-And-Match for NLVG that seamlessly unifies the strengths of two streams of NLVG methods: proposal-free and proposal-based; the former explores the search space to find time segments directly, and the latter matches the predefined time segments with ground truths. To achieve this, we formulate NLVG as a set prediction problem and design an end-to-end trainable Language Video Transformer (LVTR) that can enjoy two favorable properties, which are rich contextualization power and parallel decoding. We train LVTR with two losses. First, temporal localization loss allows time segments of all queries to regress targets (explore). Second, set guidance loss couples every query with their respective target (match). To our surprise, we found that training schedule shows divide-and-conquer-like pattern: time segments are first diversified regardless of the target, then coupled with each target, and fine-tuned to the target again. Moreover, LVTR is highly efficient and effective: it infers faster than previous baselines (by 2X or more) and sets competitive results on two NLVG benchmarks (ActivityCaptions and Charades-STA). Codes are available at https://github.com/sangminwoo/Explore-And-Match., Comment: Code: https://github.com/sangminwoo/Explore-And-Match

Online action detection, which aims to identify an ongoing action from a streaming video, is an important subject in real-world applications. For this task, previous methods use recurrent neural networks for modeling temporal relations in an input sequence. However, these methods overlook the fact that the input image sequence includes not only the action of interest but background and irrelevant actions. This would induce recurrent units to accumulate unnecessary information for encoding features on the action of interest. To overcome this problem, we propose a novel recurrent unit, named Information Discrimination Unit (IDU), which explicitly discriminates the information relevancy between an ongoing action and others to decide whether to accumulate the input information. This enables learning more discriminative representations for identifying an ongoing action. In this paper, we further present a new recurrent unit, called Information Integration Unit (IIU), for action anticipation. Our IIU exploits the outputs from IDU as pseudo action labels as well as RGB frames to learn enriched features of observed actions effectively. In experiments on TVSeries and THUMOS-14, the proposed methods outperform state-of-the-art methods by a significant margin in online action detection and action anticipation. Moreover, we demonstrate the effectiveness of the proposed units by conducting comprehensive ablation studies., Comment: To appear in TPAMI. arXiv admin note: substantial text overlap with arXiv:1912.04461

We represent three-dimensional Fourier transform light scattering, a method to reconstruct angle-resolved light scattering (ARLS) with extended angle-range from individual spherical objects. To overcome the angle limitation determined by the physical numerical aperture of an optical system, the optical light fields scattered from a sample are measured with various illumination angles, and then synthesized onto the Ewald Sphere corresponding to the normal illumination in Fourier space by rotating the scattered light signals. The method extends the angle range of the ARLS spectra beyond 90 degree, beyond the limit of forward optical measurements. Extended scattered light fields in 3D and corresponding ARLS spectra of individual microscopic polystyrene beads, and protein droplets are represented.

Visible-infrared person re-identification (VI-ReID) is an important task in night-time surveillance applications, since visible cameras are difficult to capture valid appearance information under poor illumination conditions. Compared to traditional person re-identification that handles only the intra-modality discrepancy, VI-ReID suffers from additional cross-modality discrepancy caused by different types of imaging systems. To reduce both intra- and cross-modality discrepancies, we propose a Hierarchical Cross-Modality Disentanglement (Hi-CMD) method, which automatically disentangles ID-discriminative factors and ID-excluded factors from visible-thermal images. We only use ID-discriminative factors for robust cross-modality matching without ID-excluded factors such as pose or illumination. To implement our approach, we introduce an ID-preserving person image generation network and a hierarchical feature learning module. Our generation network learns the disentangled representation by generating a new cross-modality image with different poses and illuminations while preserving a person's identity. At the same time, the feature learning module enables our model to explicitly extract the common ID-discriminative characteristic between visible-infrared images. Extensive experimental results demonstrate that our method outperforms the state-of-the-art methods on two VI-ReID datasets. The source code is available at: https://github.com/bismex/HiCMD., Comment: Accepted to CVPR 2020

Recent temporal action proposal generation approaches have suggested integrating segment- and snippet score-based methodologies to produce proposals with high recall and accurate boundaries. In this paper, different from such a hybrid strategy, we focus on the potential of the snippet score-based approach. Specifically, we propose a new snippet score-based method, named Snippet Relatedness-based Generator (SRG), with a novel concept of "snippet relatedness". Snippet relatedness represents which snippets are related to a specific action instance. To effectively learn this snippet relatedness, we present "pyramid non-local operations" for locally and globally capturing long-range dependencies among snippets. By employing these components, SRG first produces a 2D relatedness score map that enables the generation of various temporal intervals reliably covering most action instances with high overlap. Then, SRG evaluates the action confidence scores of these temporal intervals and refines their boundaries to obtain temporal action proposals. On THUMOS-14 and ActivityNet-1.3 datasets, SRG outperforms state-of-the-art methods for temporal action proposal generation. Furthermore, compared to competing proposal generators, SRG leads to significant improvements in temporal action detection., Comment: To appear in TCSVT

Most video person re-identification (re-ID) methods are mainly based on supervised learning, which requires cross-camera ID labeling. Since the cost of labeling increases dramatically as the number of cameras increases, it is difficult to apply the re-identification algorithm to a large camera network. In this paper, we address the scalability issue by presenting deep representation learning without ID information across multiple cameras. Technically, we train neural networks to generate both ID-discriminative and camera-invariant features. To achieve the ID discrimination ability of the embedding features, we maximize feature distances between different person IDs within a camera by using a metric learning approach. At the same time, considering each camera as a different domain, we apply adversarial learning across multiple camera domains for generating camera-invariant features. We also propose a part-aware adaptation module, which effectively performs multi-camera domain invariant feature learning in different spatial regions. We carry out comprehensive experiments on three public re-ID datasets (i.e., PRID-2011, iLIDS-VID, and MARS). Our method outperforms state-of-the-art methods by a large margin of about 20\% in terms of rank-1 accuracy on the large-scale MARS dataset.

Detecting fashion landmarks is a fundamental technique for visual clothing analysis. Due to the large variation and non-rigid deformation of clothes, localizing fashion landmarks suffers from large spatial variances across poses, scales, and styles. Therefore, understanding contextual knowledge of clothes is required for accurate landmark detection. To that end, in this paper, we propose a fashion landmark detection network with a global-local embedding module. The global-local embedding module is based on a non-local operation for capturing long-range dependencies and a subsequent convolution operation for adopting local neighborhood relations. With this processing, the network can consider both global and local contextual knowledge for a clothing image. We demonstrate that our proposed method has an excellent ability to learn advanced deep feature representations for fashion landmark detection. Experimental results on two benchmark datasets show that the proposed network outperforms the state-of-the-art methods. Our code is available at https://github.com/shumming/GLE_FLD., Comment: Accepted to ICCV 2019 Workshop: Computer Vision for Fashion, Art and Design