7 skills found
yule-BUAA / DNNTSPcodes of DNNTSP model for Temporal Sets Prediction
HaojiHu / Sets2SetsSequential sets to sequential sets learning
vineeths96 / Video Frame PredictionIn this repository, we focus on video frame prediction the task of predicting future frames given a set of past frames. We present an Adversarial Spatio-Temporal Convolutional LSTM architecture to predict the future frames of the Moving MNIST Dataset. We evaluate the model on long-term future frame prediction and its performance of the model on out-of-domain inputs by providing sequences on which the model was not trained.
urmilkadakia / Rainfall Prediction For The State Of Gujarat Using Deep Learning TechniquePrediction of rainfall which varies both spatially and temporally is extremely challenging. Infrared and visible spectral data from satellites have been extensively used for rainfall prediction. In this study, two deep learning methods MLP and LSTM are discussed at length for predicting precipitation at a fine spatial (10km × 10km) and temporal (hourly) resolution for the state of Gujarat. These methods are applied by using the multispectral (VIS, SWIR, MIR, WV, TIR1, TIR2) channel data such as cloud top temperature and radiance values of the INSAT-3D satellite (ISRO) as features for the model. Textural features of satellite images are incorporated by considering mean and standard deviation of each pixel’s neighbourhood. Rainfall also heavily depends on the elevation and vegetation of earth’s surface so we have used SRTM DEM and AWIFS NDVI respectively. Measurements of actual rainfall are obtained from AWS (point source stations) and TRMM (10km × 10km resolution). First dataset contains only TIR1 band temperature and AWS rainfall data for training but the second dataset includes multispectral channel data and TRMM rainfall data which brought about great improvement in results. For each data- set, a comparison between MLP and LSTM models is discussed here. We were able to classify the rainfall into nil (0mm), low ( < 2mm), medium ( > = 2mm and < 5mm) and high ( > = 5 mm) with a high accuracy. Metrics like accuracy, precision, recall and fscore have been computed to get better insights about the dataset and its corresponding outcome. Our results show that LSTM performs significantly better than MLP for any given balanced class data-sets.
yule-BUAA / ETGNNcodes of Element-guided Temporal Graph Representation Learning for Temporal Sets Prediction
easonwhite928 / DSNTSPCodes for our SIGIR 2020 paper "Dual Sequential Network for Temporal Sets Prediction"
ITMO-NSS-team / Sea Ice TransformersThis repository contains code for the research of transformer effectiveness for spatio-temporal forecasting task in comparison with 3d and 2d CNN models. The experiments was set up for sea ice concentration long-term prediction in Arctic seas.
