Search arXiv⌕ Search

arXiv subjects

Wang

Publications and source records attributed to Wang.

59 records · Page 4Linked to original sources
arXiv2018

Future Semantic Segmentation with Convolutional LSTM

We consider the problem of predicting semantic segmentation of future frames in a video. Given several observed frames in a video, our goal is to predict the semantic segmentation map of future frames that are not yet observed. A reliable solution to this problem is useful in many applications that require real-time decision making, such as autonomous driving. We propose a novel model that uses convolutional LSTM (ConvLSTM) to encode the spatiotemporal information of observed frames for future prediction. We also extend our model to use bidirectional ConvLSTM to capture temporal information in both directions. Our proposed approach outperforms other state-of-the-art methods on the benchmark dataset.

cs.CV↗
arXiv2017

MiniCLEAN Dark Matter Experiment

The MiniCLEAN dark matter experiment will exploit a single-phase liquid-argon detector instrumented with 92 photomultiplier tubes placed in the cryogen temperature with 4-$π$ coverage of a 500 kg (150 kg) target (fiducial) mass. The detector design strategy emphasizes scalability to target masses of order 10 tons or more. The detector is designed also for a liquid neon target that allows for an independent verification of signal and background and a test of the expected dependence of the WIMP-nucleus interaction rate. For MiniCLEAN, PMT stability and calibration are essential. The \textit{In-situ} optical calibration will be able monitor the PMT stability and maintain the calibration. In MiniCLEAN, we use a Light-Emitting Diode(LED)- based light injection system to provide single photons for calibration, the calibration can be performed in near real-time, providing a continuous monitor at the condition of the detector. The intrinsic $^{39}$Ar beta emitter provides another way to calibrate the detector thanks to well defined properties and uniformly distributed inside the detector volume. The energy scale can be determined by fitting the energy spectrum of experimental $^{39}$Ar data. Moreover, the preliminary results from cold gas run shows the best measurement on triplet lifetime ($\sim$ 3.5 $μ$ s). The results confirms the high purity of argon is attained by MiniCLEAN's purification system. The long triplet lifetime in gaseous argon can be exploit to obtain better performance of pulse shape discrimination (PSD) for future dark matter detector, also the low density of gaseous argon reduced the multi-scattering neutron backgrounds. On the other hand, by injecting $^{39}$Ar spike, the electronic recoil events due to $^{39}$Ar beta decay can be used to test the limit of PSD in liquid argon. The results will be informative for future multi-tonne LAr detector.

physics.ins-det↗
arXiv2016

High Dimensional Data Modeling Techniques for Detection of Chemical Plumes and Anomalies in Hyperspectral Images and Movies

We briefly review recent progress in techniques for modeling and analyzing hyperspectral images and movies, in particular for detecting plumes of both known and unknown chemicals. For detecting chemicals of known spectrum, we extend the technique of using a single subspace for modeling the background to a "mixture of subspaces" model to tackle more complicated background. Furthermore, we use partial least squares regression on a resampled training set to boost performance. For the detection of unknown chemicals we view the problem as an anomaly detection problem, and use novel estimators with low-sampled complexity for intrinsically low-dimensional data in high-dimensions that enable us to model the "normal" spectra and detect anomalies. We apply these algorithms to benchmark data sets made available by the Automated Target Detection program co-funded by NSF, DTRA and NGA, and compare, when applicable, to current state-of-the-art algorithms, with favorable results.

stat.ML↗
arXiv2014

Modeling Light Curves for Improved Classification

Many synoptic surveys are observing large parts of the sky multiple times. The resulting lightcurves provide a wonderful window to the dynamic nature of the universe. However, there are many significant challenges in analyzing these light curves. These include heterogeneity of the data, irregularly sampled data, missing data, censored data, known but variable measurement errors, and most importantly, the need to classify in astronomical objects in real time using these imperfect light curves. We describe a modeling-based approach using Gaussian process regression for generating critical measures representing features for the classification of such lightcurves. We demonstrate that our approach performs better by comparing it with past methods. Finally, we provide future directions for use in sky-surveys that are getting even bigger by the day.

stat.AP↗