Youngki Lee

Hyunseok Oh, Juheon Yi, Youngki Lee

Transformer-based models recently reached state-of-the-art single-channel speech separation accuracy; However, their extreme computational load makes it difficult to deploy them in resource-constrained mobile or IoT devices. We thus present Papez, a lightweight and computation-efficient single-channel speech separation model. Papez is based on three key techniques. We first replace the inter-chunk Transformer with small-sized auditory working memory. Second, we adaptively prune the input tokens that do not need further processing. Finally, we reduce the number of parameters through the recurrent transformer. Our extensive evaluation shows that Papez achieves the best resource and accuracy tradeoffs with a large margin. We publicly share our source code at \texttt{https://github.com/snuhcs/Papez}

Hyunseok Oh, Youngki Lee

Spiking neural network (SNN) is studied in multidisciplinary domains to (i) enable order-of-magnitudes energy-efficient AI inference and (ii) computationally simulate neuro-scientific mechanisms. The lack of discrete theory obstructs the practical application of SNN by limiting its performance and nonlinearity support. We present a new optimization-theoretic perspective of the discrete dynamics of spiking neurons. We prove that a discrete dynamical system of simple integrate-and-fire models approximates the sub-gradient method over unconstrained optimization problems. We practically extend our theory to introduce a novel sign gradient descent (signGD)-based neuronal dynamics that can (i) approximate diverse nonlinearities beyond ReLU and (ii) advance ANN-to-SNN conversion performance in low time steps. Experiments on large-scale datasets show that our technique achieves (i) state-of-the-art performance in ANN-to-SNN conversion and (ii) is the first to convert new DNN architectures, e.g., ConvNext, MLP-Mixer, and ResMLP. We publicly share our source code at https://github.com/snuhcs/snn_signgd .

Junseok Park, Inwoo Hwang, Min Whoo Lee et al.

The initial years of an infant's life are known as the critical period, during which the overall development of learning performance is significantly impacted due to neural plasticity. In recent studies, an AI agent, with a deep neural network mimicking mechanisms of actual neurons, exhibited a learning period similar to human's critical period. Especially during this initial period, the appropriate stimuli play a vital role in developing learning ability. However, transforming human cognitive bias into an appropriate shaping reward is quite challenging, and prior works on critical period do not focus on finding the appropriate stimulus. To take a step further, we propose multi-stage reinforcement learning to emphasize finding ``appropriate stimulus" around the critical period. Inspired by humans' early cognitive-developmental stage, we use multi-stage guidance near the critical period, and demonstrate the appropriate shaping reward (stage-2 guidance) in terms of the AI agent's performance, efficiency, and stability.

Taeho Kang, Kyungjin Lee, Jinrui Zhang et al.

We present Ego3DPose, a highly accurate binocular egocentric 3D pose reconstruction system. The binocular egocentric setup offers practicality and usefulness in various applications, however, it remains largely under-explored. It has been suffering from low pose estimation accuracy due to viewing distortion, severe self-occlusion, and limited field-of-view of the joints in egocentric 2D images. Here, we notice that two important 3D cues, stereo correspondences, and perspective, contained in the egocentric binocular input are neglected. Current methods heavily rely on 2D image features, implicitly learning 3D information, which introduces biases towards commonly observed motions and leads to low overall accuracy. We observe that they not only fail in challenging occlusion cases but also in estimating visible joint positions. To address these challenges, we propose two novel approaches. First, we design a two-path network architecture with a path that estimates pose per limb independently with its binocular heatmaps. Without full-body information provided, it alleviates bias toward trained full-body distribution. Second, we leverage the egocentric view of body limbs, which exhibits strong perspective variance (e.g., a significantly large-size hand when it is close to the camera). We propose a new perspective-aware representation using trigonometry, enabling the network to estimate the 3D orientation of limbs. Finally, we develop an end-to-end pose reconstruction network that synergizes both techniques. Our comprehensive evaluations demonstrate that Ego3DPose outperforms state-of-the-art models by a pose estimation error (i.e., MPJPE) reduction of 23.1% in the UnrealEgo dataset. Our qualitative results highlight the superiority of our approach across a range of scenarios and challenges.

Taeho Kang, Youngki Lee

We present EgoTAP, a heatmap-to-3D pose lifting method for highly accurate stereo egocentric 3D pose estimation. Severe self-occlusion and out-of-view limbs in egocentric camera views make accurate pose estimation a challenging problem. To address the challenge, prior methods employ joint heatmaps-probabilistic 2D representations of the body pose, but heatmap-to-3D pose conversion still remains an inaccurate process. We propose a novel heatmap-to-3D lifting method composed of the Grid ViT Encoder and the Propagation Network. The Grid ViT Encoder summarizes joint heatmaps into effective feature embedding using self-attention. Then, the Propagation Network estimates the 3D pose by utilizing skeletal information to better estimate the position of obscure joints. Our method significantly outperforms the previous state-of-the-art qualitatively and quantitatively demonstrated by a 23.9\% reduction of error in an MPJPE metric. Our source code is available in GitHub.

Taeho Kang, Jaeyeon Park, Kyungjin Lee et al.

Existing 4D Gaussian Splatting (4DGS) methods struggle to accurately reconstruct dynamic scenes, often failing to resolve ambiguous pixel correspondences and inadequate densification in dynamic regions. We address these issues by introducing a novel method composed of two key components: (1) Elliptical Error Clustering and Error Correcting Splat Addition that pinpoints dynamic areas to improve and initialize fitting splats, and (2) Grouped 4D Gaussian Splatting that improves consistency of mapping between splats and represented dynamic objects. Specifically, we classify rendering errors into missing-color and occlusion types, then apply targeted corrections via backprojection or foreground splitting guided by cross-view color consistency. Evaluations on Neural 3D Video and Technicolor datasets demonstrate that our approach significantly improves temporal consistency and achieves state-of-the-art perceptual rendering quality, improving 0.39dB of PSNR on the Technicolor Light Field dataset. Our visualization shows improved alignment between splats and dynamic objects, and the error correction method's capability to identify errors and properly initialize new splats. Our implementation details and source code are available at https://github.com/tho-kn/cem-4dgs.

Changmin Jeon, Seonjun Kim, Juheon Yi et al.

In this paper, we present Mondrian, an edge system that enables high-performance object detection on high-resolution video streams. Many lightweight models and system optimization techniques have been proposed for resource-constrained devices, but they do not fully utilize the potential of the accelerators over dynamic, high-resolution videos. To enable such capability, we devise a novel Compressive Packed Inference to minimize per-pixel processing costs by selectively determining the necessary pixels to process and combining them to maximize processing parallelism. In particular, our system quickly extracts ROIs and dynamically shrinks them, reflecting the effect of the fast-changing characteristics of objects and scenes. It then intelligently combines such scaled ROIs into large canvases to maximize the utilization of inference accelerators such as GPU. Evaluation across various datasets, models, and devices shows Mondrian outperforms state-of-the-art baselines (e.g., input rescaling, ROI extractions, ROI extractions+batching) by 15.0-19.7% higher accuracy, leading to $\times$6.65 higher throughput than frame-wise inference for processing various 1080p video streams. We will release the code after the paper review.

Hyunseok Oh, Sam Stern, Youngki Lee et al.

Natural language understanding requires interleaving textual and logical reasoning, yet large language models often fail to perform such reasoning reliably. Existing neurosymbolic systems combine LLMs with solvers but remain limited to fully formalizable tasks such as math or program synthesis, leaving natural documents with only partial logical structure unaddressed. We introduce Logitext, a neurosymbolic language that represents documents as natural language text constraints (NLTCs), making partial logical structure explicit. We develop an algorithm that integrates LLM-based constraint evaluation with satisfiability modulo theory (SMT) solving, enabling joint textual-logical reasoning. Experiments on a new content moderation benchmark, together with LegalBench and Super-Natural Instructions, show that Logitext improves both accuracy and coverage. This work is the first that treats LLM-based reasoning as an SMT theory, extending neurosymbolic methods beyond fully formalizable domains.

Kichang Yang, Seonjun Kim, Minjae Kim et al.

Edge deployment of large Vision-Language Models (VLMs) increasingly relies on flash-based weight offloading, where activation sparsification is used to reduce I/O overhead. However, conventional sparsification remains model-centric, selecting neurons solely by activation magnitude and neglecting how access patterns influence flash performance. We present Neuron Chunking, an I/O-efficient sparsification strategy that operates on chunks (i.e., groups of contiguous neurons in memory) and couples neuron importance with storage access cost. The method models I/O latency through a lightweight abstraction of access contiguity and selects chunks with high utility, defined as neuron importance normalized by estimated latency. By aligning sparsification decisions with the underlying storage behavior, Neuron Chunking improves I/O efficiency by up to 4.65x and 5.76x on Jetson Orin Nano and Jetson AGX Orin, respectively.

Jinrui Zhang, Deyu Zhang, Tingting Long et al.

We present MobiFuse, a high-precision depth perception system on mobile devices that combines dual RGB and Time-of-Flight (ToF) cameras. To achieve this, we leverage physical principles from various environmental factors to propose the Depth Error Indication (DEI) modality, characterizing the depth error of ToF and stereo-matching. Furthermore, we employ a progressive fusion strategy, merging geometric features from ToF and stereo depth maps with depth error features from the DEI modality to create precise depth maps. Additionally, we create a new ToF-Stereo depth dataset, RealToF, to train and validate our model. Our experiments demonstrate that MobiFuse excels over baselines by significantly reducing depth measurement errors by up to 77.7%. It also showcases strong generalization across diverse datasets and proves effectiveness in two downstream tasks: 3D reconstruction and 3D segmentation. The demo video of MobiFuse in real-life scenarios is available at the de-identified YouTube link(https://youtu.be/jy-Sp7T1LVs).

Taeahn Kwon, Minkyung Jeong, Eon-Suk Ko et al.

To acquire language, children need rich language input. However, many parents find it difficult to provide children with sufficient language input, which risks delaying their language development. To aid these parents, we design Captivate!, the first system that provides contextual language guidance to parents during play. Our system tracks both visual and spoken language cues to infer targets of joint attention, enabling the real-time suggestion of situation-relevant phrases for the parent. We design our system through a user-centered process with immigrant families--a highly vulnerable yet understudied population--as well as professional speech language therapists. Next, we evaluate Captivate! on parents with children aged 1-3 to observe improvements in responsive language use. We share insights into developing contextual guidance technology for linguistically diverse families.

Junseok Park, Kwanyoung Park, Hyunseok Oh et al.

Critical periods are phases during which a toddler's brain develops in spurts. To promote children's cognitive development, proper guidance is critical in this stage. However, it is not clear whether such a critical period also exists for the training of AI agents. Similar to human toddlers, well-timed guidance and multimodal interactions might significantly enhance the training efficiency of AI agents as well. To validate this hypothesis, we adapt this notion of critical periods to learning in AI agents and investigate the critical period in the virtual environment for AI agents. We formalize the critical period and Toddler-guidance learning in the reinforcement learning (RL) framework. Then, we built up a toddler-like environment with VECA toolkit to mimic human toddlers' learning characteristics. We study three discrete levels of mutual interaction: weak-mentor guidance (sparse reward), moderate mentor guidance (helper-reward), and mentor demonstration (behavioral cloning). We also introduce the EAVE dataset consisting of 30,000 real-world images to fully reflect the toddler's viewpoint. We evaluate the impact of critical periods on AI agents from two perspectives: how and when they are guided best in both uni- and multimodal learning. Our experimental results show that both uni- and multimodal agents with moderate mentor guidance and critical period on 1 million and 2 million training steps show a noticeable improvement. We validate these results with transfer learning on the EAVE dataset and find the performance advancement on the same critical period and the guidance.

Kwanyoung Park, Hyunseok Oh, Youngki Lee

Building human-like agent, which aims to learn and think like human intelligence, has long been an important research topic in AI. To train and test human-like agents, we need an environment that imposes the agent to rich multimodal perception and allows comprehensive interactions for the agent, while also easily extensible to develop custom tasks. However, existing approaches do not support comprehensive interaction with the environment or lack variety in modalities. Also, most of the approaches are difficult or even impossible to implement custom tasks. In this paper, we propose a novel VR-based toolkit, VECA, which enables building fruitful virtual environments to train and test human-like agents. In particular, VECA provides a humanoid agent and an environment manager, enabling the agent to receive rich human-like perception and perform comprehensive interactions. To motivate VECA, we also provide 24 interactive tasks, which represent (but are not limited to) four essential aspects in early human development: joint-level locomotion and control, understanding contexts of objects, multimodal learning, and multi-agent learning. To show the usefulness of VECA on training and testing human-like learning agents, we conduct experiments on VECA and show that users can build challenging tasks for engaging human-like algorithms, and the features supported by VECA are critical on training human-like agents.

Kwanyoung Park, Junseok Park, Hyunseok Oh et al.

One of the inherent limitations of current AI systems, stemming from the passive learning mechanisms (e.g., supervised learning), is that they perform well on labeled datasets but cannot deduce knowledge on their own. To tackle this problem, we derive inspiration from a highly intentional learning system via action: the toddler. Inspired by the toddler's learning procedure, we design an interactive agent that can learn and store task-agnostic visual representation while exploring and interacting with objects in the virtual environment. Experimental results show that such obtained representation was expandable to various vision tasks such as image classification, object localization, and distance estimation tasks. In specific, the proposed model achieved 100%, 75.1% accuracy and 1.62% relative error, respectively, which is noticeably better than autoencoder-based model (99.7%, 66.1%, 1.95%), and also comparable with those of supervised models (100%, 87.3%, 0.71%).

Jagmohan Chauhan, Suranga Seneviratne, Yining Hu et al.

Recurrent neural networks (RNNs) have shown promising results in audio and speech processing applications due to their strong capabilities in modelling sequential data. In many applications, RNNs tend to outperform conventional models based on GMM/UBMs and i-vectors. Increasing popularity of IoT devices makes a strong case for implementing RNN based inferences for applications such as acoustics based authentication, voice commands, and edge analytics for smart homes. Nonetheless, the feasibility and performance of RNN based inferences on resources-constrained IoT devices remain largely unexplored. In this paper, we investigate the feasibility of using RNNs for an end-to-end authentication system based on breathing acoustics. We evaluate the performance of RNN models on three types of devices; smartphone, smartwatch, and Raspberry Pi and show that unlike CNN models, RNN models can be easily ported onto resource-constrained devices without a significant loss in accuracy.