Accurate mapping of crop production is vital for food security, resource management, and sustainable agriculture. Deep learning models, especially CNNs, have shown promise in an-alyzing high-resolution satellite imagery to identify the types of crops being grown in different regions. Crop mapping is essentially a task of satellite image segmentation, for which U-Net is a highly successful family. However, deep CNNs involve more parameters and hence require heavier computations. To address this, we propose integrating the Asymmetric Convolution technique into the UNet architecture. We curated a dataset for crop production mapping using USGS Landsat 8 data and USDA NASS Cropland Data Layer. By incorporating Asymmetric Convolution into UNet, we found improved accuracy, and efficiency. Our proposed AsymmUNet model demonstrated superior performance compared to existing state-of-the-art models, such as U-Net, UNet++, and HRNet, with reduced computational complexity. Hence this approach furthers the idea of GreenAI and Sustainable Computing while solving a task that is relevant to the Sustainable Development Goals.

The Sustainable Development Goals of the UN include inclusive, safe, resilient, and sustainable human settlements. However, many people still live in substandard conditions in urban slums. To improve their conditions, the first step is to map the slum locations. One solution approach is by using Machine Learning models on Remote Sensing imagery. This study examines a Deep Learning model, the U-Net model, for slum delineation in five cities in Maharashtra, India. Our goal is to categorise localities in the five cities as slums or non-slums. Sentinel-2 satellite imagery (having a spatial resolution of 10m) is used as input for the year 2018, utilising bands 2, 3, 4, and 8. QGIS's digitization and visual interpretation tools are used to generate the labels. The model is trained on one city and tested on others, with an accuracy of 95%- 99%. Furthermore, we investigate the role of model parameters like the patch size.

Visual surveillance has become indispensable in the evolution of Intelligent Transportation Systems (ITS). Video object trajectories are key to many of the visual surveillance applications. Classifying varying length time series data such as video object trajectories using conventional neural networks, can be challenging. In this paper, we propose trajectory classification and anomaly detection using a hybrid Convolutional Neural Network (CNN) and Variational Autoencoder (VAE) architecture. First, we introduce a high level features for varying length object trajectories using color gradient representation. In the next stage, a semi-supervised way to annotate moving object trajectories extracted using Temporally Incremental Gravitational Model (TIGM) is used for class labeling. For training, anomalous trajectories are identified using t-Distributed Stochastic Neighbor Embedding (t-SNE). Finally, a hybrid CNN-VAE architecture has been proposed for trajectory classification and anomaly detection. The results obtained using publicly available surveillance video datasets reveal that the proposed method can successfully identify traffic anomalies such as violations in lane driving, sudden speed variations, abrupt termination of vehicle during movement, and vehicles moving in wrong directions. The accuracy of trajectory classification improves by a margin of 1-6% against popular neural networks-based classifiers across various datasets using the proposed high-level features. The gradient representation also improves the anomaly detection accuracy significantly (30-35%). Code and dataset can be found athttps://github.com/santhoshkelathodi/CNN-VAE.

Many physical processes involve spatio-temporal observations, which can be studied at different spatial and temporal scales. For example, rainfall data measured daily by rain gauges can be considered at daily, monthly or annual temporal scales, and local, grid-wise, region-wise or country-wise spatial scales. In this work, we focus on detection of anomalies in such multi-scale spatio-temporal data. We consider an anomaly as an event where the measured values over a spatio-temporally extended region are significantly different from their long-term means. However we aim to avoid setting any thresholds on the measured values and spatio-temporal sizes, because not only are thresholds subjective but also the long-term mean values often vary spatially and temporally. For this purpose we use spatio-Temporal Markov Random Field, where latent states indicate anomaly type (positive anomaly, negative anomaly, no anomaly/normal). Spatio-temporal coherence is maintained through suitable edge potentials. The model is extended to multiple spatio-temporal scales to achieve our second goal: anomalies at any scale should be defined both on the data at that scale, and also on anomalies at other scales. This allows us to trace an anomaly at a coarse scale to finer scales. For example, whether rainfall in a particular year is anomalous over a region should depend not only on the total volume of rainfall over the entire region, but also on whether there were such anomalies at the grid-scale, and the monthly scale. We use this approach to study rainfall anomalies over India-extremely diverse with respect to rainfall-for the period 1901-2011, and show its benefits over existing approaches.

Simulation of rainfall over a region for long time-sequences can be very useful for planning and policy-making, especially in India where the economy is heavily reliant on monsoon rainfall. However, such simulations should be able to preserve known spatial and temporal characteristics of rainfall over India. General Circulation Models (GCMs) are unable to do so, and various rainfall generators designed by hydrologists using stochastic processes like Gaussian Processes are also difficult to apply over the highly diverse landscape of India. In this paper, we explore a series of Bayesian models based on conditional distributions of latent variables that describe weather conditions at specific locations and over the whole country. During parameter estimation from observed data, we use spatio-temporal smoothing using Markov Random Field so that the parameters learnt are spatially and temporally coherent. Also, we use a nonparametric spatial clustering based on Chinese Restaurant Process to identify homogeneous regions, which are utilized by some of the proposed models to improve spatial correlations of the simulated rainfall. The models are able to simulate daily rainfall across India for years, and can also utilize contextual information for conditional simulation. We use two datasets of different spatial resolutions over India, and focus on the period 2000--2015. We consider metrics to study the spatio-temporal properties of the simulations by the models, and compare them with the observed data to evaluate the strengths and weaknesses of the models.

A video is understood by users in terms of entities present in it. Entity Discovery is the task of building appearance model for each entity (e.g., a person), and finding all its occurrences in the video. We represent a video as a sequence of tracklets, each spanning 10-20 frames, and associated with one entity. We pose Entity Discovery as tracklet clustering, and approach it by leveraging Temporal Coherence (TC): the property that temporally neighboring tracklets are likely to be associated with the same entity. Our major contributions are the first Bayesian nonparametric models for TC at tracklet-level. We extend Chinese Restaurant Process (CRP) to TC-CRP, and further to Temporally Coherent Chinese Restaurant Franchise (TC-CRF) to jointly model entities and temporal segments using mixture components and sparse distributions. For discovering persons in TV serial videos without meta-data like scripts, these methods show considerable improvement over state-of-the-art approaches to tracklet clustering in terms of clustering accuracy, cluster purity and entity coverage. The proposed methods can perform online tracklet clustering on streaming videos unlike existing approaches, and can automatically reject false tracklets. Finally we discuss entity-driven video summarization- where temporal segments of the video are selected based on the discovered entities, to create a semantically meaningful summary.

Tracklet Clustering is central to several Computer vision tasks [17][20]. A video can be represented as a sequence of tracklets, each spanning over 10–20 successive video frames, and each tracklet is associated with one entity (eg. person in case of TV-serial videos). Tracklets are instances of data-types exhibiting rich spatio-temporal structure. Existing approaches model tracklets by deploying detailed parametric models with a large number of parameters, making the inference unwieldy. The task of Person Discovery in long TV-series videos (40–45 minutes) with many persons can be naturally posed as tracklet clustering, and existing approaches give unsatisfactory performance on it. In this paper we attempt to leverage Temporal Coherence(TC) of videos to improve tracklet clustering. TC is the fundamental property of videos that each tracklet is likely to be associated with the same entity as its predecessor or successor. We propose the first Bayesian nonparametric approach for modelling TC, which can automatically infer the number of clusters to be formed. The major contribution of this paper is Temporally Coherent Chinese Restaurant Process (TC-CRP), which extends CRP by using TC. On the task of discovering persons in TV serials via tracklet clustering, without meta-data such as scripts, TC-CRP shows up to 25% improvement in cluster purity compared to state-of-the-art parametric models, and upto 36% improvement in number of persons discovered. We use a simple representation of tracklets: a vector of very generic features (like pixel intensity) which can correspond to any type of entity (not necessarily person), and empirically demonstrate the utility of TC-CRP for discovering entities like cars and planes. Moreover, unlike existing approaches TC-CRP can perform online tracklet clustering on streaming videos with very little performance deterioration, and can also automatically reject outliers (tracklets resulting from false detections).

Reconciling opinions from multiple sources on questions of interest to determine the correct answers is an important problem encountered in collaborative information systems such as Q & A forums and prediction markets. Our current work focuses on a widely applicable variant of the above problem where the opinions and answers are categorical-valued with the set of values possibly varying across questions. Most of the existing techniques are tailored only for binary opinions and cannot be effectively adapted for questions with categorical opinions. To address this, we propose a generic Bayesian framework for opinion reconciliation that can readily incorporate latent and observed attributes of sources and subjects. For the scenario of interest, we derive three specific model instantiations of the general approach (CTM, CTM-OSF, CTM-LSG), which respectively capture the latent source behavior, variations of source behavior across subject groups, and inter-source correlations. Empirical results on real-world datasets point to the relative superiority of the proposed models over existing baselines.

The highly dynamic nature of online commenting environments makes accurate ratings prediction for new comments challenging. In such a setting, in addition to exploiting comments with high predicted ratings, it is also critical to explore comments with high uncertainty in the predictions. In this paper, we propose a novel upper confidence bound (UCB) algorithm called LOGUCB that balances exploration with exploitation when the average rating of a comment is modeled using logistic regression on its features. At the core of our LOGUCB algorithm lies a novel variance approximation technique for the Bayesian logistic regression model that is used to compute the UCB value for each comment. In experiments with a real-life comments dataset from Yahoo! News, we show that LOGUCB with bag-of-words and topic features outperforms state-of-the-art explore-exploit algorithms.

We propose a novel space-time descriptor for region-based tracking which is very concise and efficient. The regions represented by covariance matrices within a temporal fragment, are used to estimate this space-time descriptor which we call the Eigenprofiles(EP). EP so obtained is used in estimating the Covariance Matrix of features over spatio-temporal fragments. The Second Order Statistics of spatio-temporal fragments form our target model which can be adapted for variations across the video. The model being concise also allows the use of multiple spatially overlapping fragments to represent the target. We demonstrate good tracking results on very challenging datasets, shot under insufficient illumination conditions.