Junghyun Cho

JunYong Choi, SeokYeong Lee, Haesol Park et al.

We propose a scene-level inverse rendering framework that uses multi-view images to decompose the scene into geometry, a SVBRDF, and 3D spatially-varying lighting. Because multi-view images provide a variety of information about the scene, multi-view images in object-level inverse rendering have been taken for granted. However, owing to the absence of multi-view HDR synthetic dataset, scene-level inverse rendering has mainly been studied using single-view image. We were able to successfully perform scene-level inverse rendering using multi-view images by expanding OpenRooms dataset and designing efficient pipelines to handle multi-view images, and splitting spatially-varying lighting. Our experiments show that the proposed method not only achieves better performance than single-view-based methods, but also achieves robust performance on unseen real-world scene. Also, our sophisticated 3D spatially-varying lighting volume allows for photorealistic object insertion in any 3D location.

SeokYeong Lee, JunYong Choi, Seungryong Kim et al.

In this paper, we introduce a new challenge for synthesizing novel view images in practical environments with limited input multi-view images and varying lighting conditions. Neural radiance fields (NeRF), one of the pioneering works for this task, demand an extensive set of multi-view images taken under constrained illumination, which is often unattainable in real-world settings. While some previous works have managed to synthesize novel views given images with different illumination, their performance still relies on a substantial number of input multi-view images. To address this problem, we suggest ExtremeNeRF, which utilizes multi-view albedo consistency, supported by geometric alignment. Specifically, we extract intrinsic image components that should be illumination-invariant across different views, enabling direct appearance comparison between the input and novel view under unconstrained illumination. We offer thorough experimental results for task evaluation, employing the newly created NeRF Extreme benchmark-the first in-the-wild benchmark for novel view synthesis under multiple viewing directions and varying illuminations.

JunYong Choi, SeokYeong Lee, Haesol Park et al.

In this paper, we propose a scene-level inverse rendering framework that uses multi-view images to decompose the scene into geometry, SVBRDF, and 3D spatially-varying lighting. While multi-view images have been widely used for object-level inverse rendering, scene-level inverse rendering has primarily been studied using single-view images due to the lack of a dataset containing high dynamic range multi-view images with ground-truth geometry, material, and spatially-varying lighting. To improve the quality of scene-level inverse rendering, a novel framework called Multi-view Attention Inverse Rendering (MAIR) was recently introduced. MAIR performs scene-level multi-view inverse rendering by expanding the OpenRooms dataset, designing efficient pipelines to handle multi-view images, and splitting spatially-varying lighting. Although MAIR showed impressive results, its lighting representation is fixed to spherical Gaussians, which limits its ability to render images realistically. Consequently, MAIR cannot be directly used in applications such as material editing. Moreover, its multi-view aggregation networks have difficulties extracting rich features because they only focus on the mean and variance between multi-view features. In this paper, we propose its extended version, called MAIR++. MAIR++ addresses the aforementioned limitations by introducing an implicit lighting representation that accurately captures the lighting conditions of an image while facilitating realistic rendering. Furthermore, we design a directional attention-based multi-view aggregation network to infer more intricate relationships between views. Experimental results show that MAIR++ not only achieves better performance than MAIR and single-view-based methods, but also displays robust performance on unseen real-world scenes.

Sithu Aung, Min-Cheol Sagong, Junghyun Cho

We address an advanced challenge of predicting pedestrian occupancy as an extension of multi-view pedestrian detection in urban traffic. To support this, we have created a new synthetic dataset called MVP-Occ, designed for dense pedestrian scenarios in large-scale scenes. Our dataset provides detailed representations of pedestrians using voxel structures, accompanied by rich semantic scene understanding labels, facilitating visual navigation and insights into pedestrian spatial information. Furthermore, we present a robust baseline model, termed OmniOcc, capable of predicting both the voxel occupancy state and panoptic labels for the entire scene from multi-view images. Through in-depth analysis, we identify and evaluate the key elements of our proposed model, highlighting their specific contributions and importance.

Minsoo Kim, Min-Cheol Sagong, Gi Pyo Nam et al.

Deep learning-based face recognition continues to face challenges due to its reliance on huge datasets obtained from web crawling, which can be costly to gather and raise significant real-world privacy concerns. To address this issue, we propose VIGFace, a novel framework capable of generating synthetic facial images. Our idea originates from pre-assigning virtual identities in the feature space. Initially, we train the face recognition model using a real face dataset and create a feature space for both real and virtual identities, where virtual prototypes are orthogonal to other prototypes. Subsequently, we train the diffusion model based on the established feature space, enabling it to generate authentic human face images from real prototypes and synthesize virtual face images from virtual prototypes. Our proposed framework provides two significant benefits. Firstly, it shows clear separability between existing individuals and virtual face images, allowing one to create synthetic images with confidence and without concerns about privacy and portrait rights. Secondly, it ensures improved performance through data augmentation by incorporating real existing images. Extensive experiments demonstrate the superiority of our virtual face dataset and framework, outperforming the previous state-of-the-art on various face recognition benchmarks.

Seokyeong Lee, Sithu Aung, Junyong Choi et al.

Monocular 3D object detection (M3OD) has long faced challenges due to data scarcity caused by high annotation costs and inherent 2D-to-3D ambiguity. Although various weakly supervised methods and pseudo-labeling methods have been proposed to address these issues, they are mostly limited by domain-specific learning or rely solely on shape information from a single observation. In this paper, we propose a novel pseudo-labeling framework that uses only video data and is more robust to occlusion, without requiring a multi-view setup, additional sensors, camera poses, or domain-specific training. Specifically, we explore a technique for aggregating the pseudo-LiDARs of both static and dynamic objects across temporally adjacent frames using object point tracking, enabling 3D attribute extraction in scenarios where 3D data acquisition is infeasible. Extensive experiments demonstrate that our method ensures reliable accuracy and strong scalability, making it a practical and effective solution for M3OD.

JunYong Choi, Min-Cheol Sagong, SeokYeong Lee et al.

We propose a diffusion-based inverse rendering framework that decomposes a single RGB image into geometry, material, and lighting. Inverse rendering is inherently ill-posed, making it difficult to predict a single accurate solution. To address this challenge, recent generative model-based methods aim to present a range of possible solutions. However, finding a single accurate solution and generating diverse solutions can be conflicting. In this paper, we propose a channel-wise noise scheduling approach that allows a single diffusion model architecture to achieve two conflicting objectives. The resulting two diffusion models, trained with different channel-wise noise schedules, can predict a single highly accurate solution and present multiple possible solutions. The experimental results demonstrate the superiority of our two models in terms of both diversity and accuracy, which translates to enhanced performance in downstream applications such as object insertion and material editing.

Yeji Choi, Hyunjung Park, Gi Pyo Nam et al.

In this paper, we introduce a new large-scale face database from KIST, denoted as K-FACE, and describe a novel capturing device specifically designed to obtain the data. The K-FACE database contains more than 1 million high-quality images of 1,000 subjects selected by considering the ratio of gender and age groups. It includes a variety of attributes, including 27 poses, 35 lighting conditions, three expressions, and occlusions by the combination of five types of accessories. As the K-FACE database is systematically constructed through a hemispherical capturing system with elaborate lighting control and multiple cameras, it is possible to accurately analyze the effects of factors that cause performance degradation, such as poses, lighting changes, and accessories. We consider not only the balance of external environmental factors, such as pose and lighting, but also the balance of personal characteristics such as gender and age group. The gender ratio is the same, while the age groups of subjects are uniformly distributed from the 20s to 50s for both genders. The K-FACE database can be extensively utilized in various vision tasks, such as face recognition, face frontalization, illumination normalization, face age estimation, and three-dimensional face model generation. We expect systematic diversity and uniformity of the K-FACE database to promote these research fields.

Je Hyeong Hong, Hanjo Kim, Minsoo Kim et al.

Face recognition now requires a large number of labelled masked face images in the era of this unprecedented COVID-19 pandemic. Unfortunately, the rapid spread of the virus has left us little time to prepare for such dataset in the wild. To circumvent this issue, we present a 3D model-based approach called WearMask3D for augmenting face images of various poses to the masked face counterparts. Our method proceeds by first fitting a 3D morphable model on the input image, second overlaying the mask surface onto the face model and warping the respective mask texture, and last projecting the 3D mask back to 2D. The mask texture is adapted based on the brightness and resolution of the input image. By working in 3D, our method can produce more natural masked faces of diverse poses from a single mask texture. To compare precisely between different augmentation approaches, we have constructed a dataset comprising masked and unmasked faces with labels called MFW-mini. Experimental results demonstrate WearMask3D produces more realistic masked faces, and utilizing these images for training leads to state-of-the-art recognition accuracy for masked faces.

Hochul Hwang, Cheongjae Jang, Geonwoo Park et al.

To train deep learning models for vision-based action recognition of elders' daily activities, we need large-scale activity datasets acquired under various daily living environments and conditions. However, most public datasets used in human action recognition either differ from or have limited coverage of elders' activities in many aspects, making it challenging to recognize elders' daily activities well by only utilizing existing datasets. Recently, such limitations of available datasets have actively been compensated by generating synthetic data from realistic simulation environments and using those data to train deep learning models. In this paper, based on these ideas we develop ElderSim, an action simulation platform that can generate synthetic data on elders' daily activities. For 55 kinds of frequent daily activities of the elders, ElderSim generates realistic motions of synthetic characters with various adjustable data-generating options, and provides different output modalities including RGB videos, two- and three-dimensional skeleton trajectories. We then generate KIST SynADL, a large-scale synthetic dataset of elders' activities of daily living, from ElderSim and use the data in addition to real datasets to train three state-of the-art human action recognition models. From the experiments following several newly proposed scenarios that assume different real and synthetic dataset configurations for training, we observe a noticeable performance improvement by augmenting our synthetic data. We also offer guidance with insights for the effective utilization of synthetic data to help recognize elders' daily activities.

Sunghun Joung, Seungryong Kim, Hanjae Kim et al.

Existing techniques to encode spatial invariance within deep convolutional neural networks only model 2D transformation fields. This does not account for the fact that objects in a 2D space are a projection of 3D ones, and thus they have limited ability to severe object viewpoint changes. To overcome this limitation, we introduce a learnable module, cylindrical convolutional networks (CCNs), that exploit cylindrical representation of a convolutional kernel defined in the 3D space. CCNs extract a view-specific feature through a view-specific convolutional kernel to predict object category scores at each viewpoint. With the view-specific feature, we simultaneously determine objective category and viewpoints using the proposed sinusoidal soft-argmax module. Our experiments demonstrate the effectiveness of the cylindrical convolutional networks on joint object detection and viewpoint estimation.

Youngduck Choi, Youngnam Lee, Junghyun Cho et al.

Knowledge tracing, the act of modeling a student's knowledge through learning activities, is an extensively studied problem in the field of computer-aided education. Although models with attention mechanism have outperformed traditional approaches such as Bayesian knowledge tracing and collaborative filtering, they share two limitations. Firstly, the models rely on shallow attention layers and fail to capture complex relations among exercises and responses over time. Secondly, different combinations of queries, keys and values for the self-attention layer for knowledge tracing were not extensively explored. Usual practice of using exercises and interactions (exercise-response pairs) as queries and keys/values respectively lacks empirical support. In this paper, we propose a novel Transformer based model for knowledge tracing, SAINT: Separated Self-AttentIve Neural Knowledge Tracing. SAINT has an encoder-decoder structure where exercise and response embedding sequence separately enter the encoder and the decoder respectively, which allows to stack attention layers multiple times. To the best of our knowledge, this is the first work to suggest an encoder-decoder model for knowledge tracing that applies deep self-attentive layers to exercises and responses separately. The empirical evaluations on a large-scale knowledge tracing dataset show that SAINT achieves the state-of-the-art performance in knowledge tracing with the improvement of AUC by 1.8% compared to the current state-of-the-art models.

Youngnam Lee, Dongmin Shin, HyunBin Loh et al.

Student dropout prediction provides an opportunity to improve student engagement, which maximizes the overall effectiveness of learning experiences. However, researches on student dropout were mainly conducted on school dropout or course dropout, and study session dropout in a mobile learning environment has not been considered thoroughly. In this paper, we investigate the study session dropout prediction problem in a mobile learning environment. First, we define the concept of the study session, study session dropout and study session dropout prediction task in a mobile learning environment. Based on the definitions, we propose a novel Transformer based model for predicting study session dropout, DAS: Deep Attentive Study Session Dropout Prediction in Mobile Learning Environment. DAS has an encoder-decoder structure which is composed of stacked multi-head attention and point-wise feed-forward networks. The deep attentive computations in DAS are capable of capturing complex relations among dynamic student interactions. To the best of our knowledge, this is the first attempt to investigate study session dropout in a mobile learning environment. Empirical evaluations on a large-scale dataset show that DAS achieves the best performance with a significant improvement in area under the receiver operating characteristic curve compared to baseline models.

Youngduck Choi, Youngnam Lee, Junghyun Cho et al.

Like many other domains in Artificial Intelligence (AI), there are specific tasks in the field of AI in Education (AIEd) for which labels are scarce and expensive, such as predicting exam score or review correctness. A common way of circumventing label-scarce problems is pre-training a model to learn representations of the contents of learning items. However, such methods fail to utilize the full range of student interaction data available and do not model student learning behavior. To this end, we propose Assessment Modeling, a class of fundamental pre-training tasks for general interactive educational systems. An assessment is a feature of student-system interactions which can serve as a pedagogical evaluation. Examples include the correctness and timeliness of a student's answer. Assessment Modeling is the prediction of assessments conditioned on the surrounding context of interactions. Although it is natural to pre-train on interactive features available in large amounts, limiting the prediction targets to assessments focuses the tasks' relevance to the label-scarce educational problems and reduces less-relevant noise. While the effectiveness of different combinations of assessments is open for exploration, we suggest Assessment Modeling as a first-order guiding principle for selecting proper pre-training tasks for label-scarce educational problems.

Youngduck Choi, Youngnam Lee, Dongmin Shin et al.

With advances in Artificial Intelligence in Education (AIEd) and the ever-growing scale of Interactive Educational Systems (IESs), data-driven approach has become a common recipe for various tasks such as knowledge tracing and learning path recommendation. Unfortunately, collecting real students' interaction data is often challenging, which results in the lack of public large-scale benchmark dataset reflecting a wide variety of student behaviors in modern IESs. Although several datasets, such as ASSISTments, Junyi Academy, Synthetic and STATICS, are publicly available and widely used, they are not large enough to leverage the full potential of state-of-the-art data-driven models and limits the recorded behaviors to question-solving activities. To this end, we introduce EdNet, a large-scale hierarchical dataset of diverse student activities collected by Santa, a multi-platform self-study solution equipped with artificial intelligence tutoring system. EdNet contains 131,441,538 interactions from 784,309 students collected over more than 2 years, which is the largest among the ITS datasets released to the public so far. Unlike existing datasets, EdNet provides a wide variety of student actions ranging from question-solving to lecture consumption and item purchasing. Also, EdNet has a hierarchical structure where the student actions are divided into 4 different levels of abstractions. The features of EdNet are domain-agnostic, allowing EdNet to be extended to different domains easily. The dataset is publicly released under Creative Commons Attribution-NonCommercial 4.0 International license for research purposes. We plan to host challenges in multiple AIEd tasks with EdNet to provide a common ground for the fair comparison between different state of the art models and encourage the development of practical and effective methods.

Youngnam Lee, Youngduck Choi, Junghyun Cho et al.

Machine learning plays an increasing role in intelligent tutoring systems as both the amount of data available and specialization among students grow. Nowadays, these systems are frequently deployed on mobile applications. Users on such mobile education platforms are dynamic, frequently being added, accessing the application with varying levels of focus, and changing while using the service. The education material itself, on the other hand, is often static and is an exhaustible resource whose use in tasks such as problem recommendation must be optimized. The ability to update user models with respect to educational material in real-time is thus essential; however, existing approaches require time-consuming re-training of user features whenever new data is added. In this paper, we introduce a neural pedagogical agent for real-time user modeling in the task of predicting user response correctness, a central task for mobile education applications. Our model, inspired by work in natural language processing on sequence modeling and machine translation, updates user features in real-time via bidirectional recurrent neural networks with an attention mechanism over embedded question-response pairs. We experiment on the mobile education application SantaTOEIC, which has 559k users, 66M response data points as well as a set of 10k study problems each expert-annotated with topic tags and gathered since 2016. Our model outperforms existing approaches over several metrics in predicting user response correctness, notably out-performing other methods on new users without large question-response histories. Additionally, our attention mechanism and annotated tag set allow us to create an interpretable education platform, with a smart review system that addresses the aforementioned issue of varied user attention and problem exhaustion.