이남훈 교수

We introduce a Deep Stochastic IOC RNN Encoder-decoder framework, DESIRE, for the task of future predictions of multiple interacting agents in dynamic scenes. DESIRE effectively predicts future locations of objects in multiple scenes by 1) accounting for the multi-modal nature of the future prediction (i.e., given the same context, future may vary), 2) foreseeing the potential future outcomes and make a strategic prediction based on that, and 3) reasoning not only from the past motion history, b

Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connecti

Pruning large neural networks while maintaining their performance is often desirable due to the reduced space and time complexity. In existing methods, pruning is done within an iterative optimization procedure with either heuristically designed pruning schedules or additional hyperparameters, undermining their utility. In this work, we present a new approach that prunes a given network once at initialization prior to training. To achieve this, we introduce a saliency criterion based on connecti

We introduce a Deep Stochastic IOC RNN Encoderdecoder framework, DESIRE, for the task of future predictions of multiple interacting agents in dynamic scenes. DESIRE effectively predicts future locations of objects in multiple scenes by 1) accounting for the multi-modal nature of the future prediction (i.e., given the same context, future may vary), 2) foreseeing the potential future outcomes and make a strategic prediction based on that, and 3) reasoning not only from the past motion history, bu

Predicting the trajectory of a wide receiver in the game of American football requires prior knowledge about the game (e.g., route trees, defensive formations) and an accurate model of how the environment will change over time (e.g., opponent reaction strategies, motion attributes of players). Our aim is to build a computational model of the wide receiver, which takes into account prior knowledge about the game and short-term predictive models of how the environment will change over time. While

In the first order decay (FOD) model of landfill methane generation, the methane generation potential ( L₀) and methane generation rate constant ( k) for both bulk municipal solid waste (MSW) and individual waste components have been determined by a variety of approaches throughout various literature. Differences in the determination methods for L₀ and k are related to differences in our understanding of the waste decomposition dynamics. A thorough understanding of the vari

Network pruning is a promising avenue for compressing deep neural networks. A typical approach to pruning starts by training a model and then removing redundant parameters while minimizing the impact on what is learned. Alternatively, a recent approach shows that pruning can be done at initialization prior to training, based on a saliency criterion called connection sensitivity. However, it remains unclear exactly why pruning an untrained, randomly initialized neural network is effective. In thi

Three novel downstream elements (GLE, DPE-L1, DPE-L2), located between +1 and +32 bp, regulate both constitutive and activated MHC class I gene expression by selectively increasing usage of transcription start sites clustered at the 5' end of the core promoter upstream of +1 bp. Results indicate that the downstream elements preferentially regulate TAF1-dependent, relative to TAF1-independent, transcription.

The aim of this study was to investigate the possibility of measuring methane emission fluxes, using surface methane concentration and gauge pressure, by analyzing the influence of gauge pressure on the methane emission flux and the surface methane concentration, as well as the correlation between the methane emission flux and surface methane concentrations. The surface methane concentration was measured using a laser methane detector. Our results show a positive linear relationship between the

Nitrous oxide (N<sub>2</sub>O) is an important greenhouse gas (GHG) and an ozone-depleting substance that can be emitted from landfills. Understanding the dynamics of N<sub>2</sub>O and its contribution to total emissions is critical to effective mitigation measures. This study outlines the dynamics of N<sub>2</sub>O in solid waste landfills and N<sub>2</sub>O emissions from them. N<sub>2</sub>O generation in anaerobic landfills is prim

This case demonstrates that long-term exposure to PTFE spraying can cause granulomatous lung lesions such as pneumoconiosis; such lesions appear to be caused not by the degradation products of PTFE from high temperatures but by spraying the particles of PTFE. Along with air-sampling analysis, we suggest monitoring the concentration of airborne PTFE particles related to chronic lung disease.

We introduce the concept of a Visual Compiler that generates a scene specific pedestrian detector and pose estimator without any pedestrian observations. Given a single image and auxiliary scene information in the form of camera parameters and geometric layout of the scene, the Visual Compiler first infers geometrically and photometrically accurate images of humans in that scene through the use of computer graphics rendering. Using these renders we learn a scene-and-region specific spatially-var