We address the problem of estimating a high quality dense depth map from a single RGB input image. We start out with a baseline encoder-decoder convolutional neural network architecture and pose the question of how the global processing of information can help improve overall depth estimation. To this end, we propose a transformerbased architecture block that divides the depth range into bins whose center value is estimated adaptively per image. The final depth values are estimated as linear combinations of the bin centers. We call our new building block AdaBins. Our results show a decisive improvement over the state-ofthe-art on several popular depth datasets across all metrics. We also validate the effectiveness of the proposed block with an ablation study and provide the code and corresponding pre-trained weights of the new state-of-the-art model 1 .

In this paper we explore the task of modeling (semi) structured object sequences; in particular we focus our attention on the problem of developing a structure-aware input representation for such sequences. In such sequences, we assume that each structured object is represented by a set of key-value pairs which encode the attributes of the structured object. Given a universe of keys, a sequence of structured objects can then be viewed as an evolution of the values for each key, over time. We encode and construct a sequential representation using the values for a particular key (Temporal Value Modeling - TVM) and then self-attend over the set of key-conditioned value sequences to a create a representation of the structured object sequence (Key Aggregation - KA). We pre-train and fine-tune the two components independently and present an innovative training schedule that interleaves the training of both modules with shared attention heads. We find that this iterative two part-training results in better performance than a unified network with hierarchical encoding as well as over, other methods that use a {\em record-view} representation of the sequence \cite{de2021transformers4rec} or a simple {\em flattened} representation of the sequence. We conduct experiments using real-world data to demonstrate the advantage of interleaving TVM-KA on multiple tasks and detailed ablation studies motivating our modeling choices. We find that our approach performs better than flattening sequence objects and also allows us to operate on significantly larger sequences than existing methods.

Before the transition of AVs to urban roads and subsequently unprecedented changes in traffic conditions, evaluation of transportation policies and futuristic road design related to pedestrian crossing behavior is of vital importance. Recent studies analyzed the non-causal impact of various variables on pedestrian waiting time in the presence of AVs. However, we mainly investigate the causal effect of traffic density on pedestrian waiting time. We develop a Double/Debiased Machine Learning (DML) model in which the impact of confounders variable influencing both a policy and an outcome of interest is addressed, resulting in unbiased policy evaluation. Furthermore, we try to analyze the effect of traffic density by developing a copula-based joint model of two main components of pedestrian crossing behavior, pedestrian stress level and waiting time. The copula approach has been widely used in the literature, for addressing self-selection problems, which can be classified as a causality analysis in travel behavior modeling. The results obtained from copula approach and DML are compared based on the effect of traffic density. In DML model structure, the standard error term of density parameter is lower than copula approach and the confidence interval is considerably more reliable. In addition, despite the similar sign of effect, the copula approach estimates the effect of traffic density lower than DML, due to the spurious effect of confounders. In short, the DML model structure can flexibly adjust the impact of confounders by using machine learning algorithms and is more reliable for planning future policies.

Dense retrievers have made significant strides in obtaining state-of-the-art results on text retrieval and open-domain question answering (ODQA). Yet most of these achievements were made possible with the help of large annotated datasets, unsupervised learning for dense retrieval models remains an open problem. In this work, we explore two categories of methods for creating pseudo query-document pairs, named query extraction (QExt) and transferred query generation (TQGen), to augment the retriever training in an annotation-free and scalable manner. Specifically, QExt extracts pseudo queries by document structures or selecting salient random spans, and TQGen utilizes generation models trained for other NLP tasks (e.g., summarization) to produce pseudo queries. Extensive experiments show that dense retrievers trained with individual augmentation methods can perform comparably well with multiple strong baselines, and combining them leads to further improvements, achieving state-of-the-art performance of unsupervised dense retrieval on both BEIR and ODQA datasets.

Neuromorphic vision or event vision is an advanced vision technology, where in contrast to the visible camera that outputs pixels, the event vision generates neuromorphic events every time there is a brightness change which exceeds a specific threshold in the field of view (FOV). This study focuses on leveraging neuromorphic event data for roadside object detection. This is a proof of concept towards building artificial intelligence (AI) based pipelines which can be used for forward perception systems for advanced vehicular applications. The focus is on building efficient state-of-the-art object detection networks with better inference results for fast-moving forward perception using an event camera. In this article, the event-simulated A2D2 dataset is manually annotated and trained on two different YOLOv5 networks (small and large variants). To further assess its robustness, single model testing and ensemble model testing are carried out.

Continuous-time Markov chains are used to model stochastic systems where transitions can occur at irregular times, e.g., birth-death processes, chemical reaction networks, population dynamics, and gene regulatory networks. We develop a method to learn a continuous-time Markov chain's transition rate functions from fully observed time series. In contrast with existing methods, our method allows for transition rates to depend nonlinearly on both state variables and external covariates. The Gillespie algorithm is used to generate trajectories of stochastic systems where propensity functions (reaction rates) are known. Our method can be viewed as the inverse: given trajectories of a stochastic reaction network, we generate estimates of the propensity functions. While previous methods used linear or log-linear methods to link transition rates to covariates, we use neural networks, increasing the capacity and potential accuracy of learned models. In the chemical context, this enables the method to learn propensity functions from non-mass-action kinetics. We test our method with synthetic data generated from a variety of systems with known transition rates. We show that our method learns these transition rates with considerably more accuracy than log-linear methods, in terms of mean absolute error between ground truth and predicted transition rates. We also demonstrate an application of our methods to open-loop control of a continuous-time Markov chain.

This paper focuses on a stochastic system identification problem: given time series observations of a stochastic differential equation (SDE) driven by L\'{e}vy $\alpha$-stable noise, estimate the SDE's drift field. For $\alpha$ in the interval $[1,2)$, the noise is heavy-tailed, leading to computational difficulties for methods that compute transition densities and/or likelihoods in physical space. We propose a Fourier space approach that centers on computing time-dependent characteristic functions, i.e., Fourier transforms of time-dependent densities. Parameterizing the unknown drift field using Fourier series, we formulate a loss consisting of the squared error between predicted and empirical characteristic functions. We minimize this loss with gradients computed via the adjoint method. For a variety of one- and two-dimensional problems, we demonstrate that this method is capable of learning drift fields in qualitative and/or quantitative agreement with ground truth fields.

Unmanned air vehicles (UAVs) popularity is on the rise as it enables the services like traffic monitoring, emergency communications, deliveries, and surveillance. However, the unauthorized usage of UAVs (a.k.a drone) may violate security and privacy protocols for security-sensitive national and international institutions. The presented challenges require fast, efficient, and precise detection of UAVs irrespective of harsh weather conditions, the presence of different objects, and their size to enable SafeSpace. Recently, there has been significant progress in using the latest deep learning models, but those models have shortcomings in terms of computational complexity, precision, and non-scalability. To overcome these limitations, we propose a precise and efficient multiscale and multifeature UAV detection network for SafeSpace, i.e., \textit{MultiFeatureNet} (\textit{MFNet}), an improved version of the popular object detection algorithm YOLOv5s. In \textit{MFNet}, we perform multiple changes in the backbone and neck of the YOLOv5s network to focus on the various small and ignored features required for accurate and fast UAV detection. To further improve the accuracy and focus on the specific situation and multiscale UAVs, we classify the \textit{MFNet} into small (S), medium (M), and large (L): these are the combinations of various size filters in the convolution and the bottleneckCSP layers, reside in the backbone and neck of the architecture. This classification helps to overcome the computational cost by training the model on a specific feature map rather than all the features. The dataset and code are available as an open source: github.com/ZeeshanKaleem/MultiFeatureNet.

Predicting emotions expressed in text is a well-studied problem in the NLP community. Recently there has been active research in extracting the cause of an emotion expressed in text. Most of the previous work has done causal emotion entailment in documents. In this work, we propose neural models to extract emotion cause span and entailment in conversations. For learning such models, we use RECCON dataset, which is annotated with cause spans at the utterance level. In particular, we propose MuTEC, an end-to-end Multi-Task learning framework for extracting emotions, emotion cause, and entailment in conversations. This is in contrast to existing baseline models that use ground truth emotions to extract the cause. MuTEC performs better than the baselines for most of the data folds provided in the dataset.

Unmanned aerial vehicles (UAVs) with on-board cameras are widely used for remote surveillance and video capturing applications. In remote virtual reality (VR) applications, multiple UAVs can be used to capture different partially overlapping angles of the ground target, which can be stitched together to provide 360{\deg} views. This requires coordinated formation of UAVs that is adaptive to movements of the ground target. In this paper, we propose a joint UAV formation and tracking framework to capture 360{\deg} angles of the target. The proposed framework uses a zero touch approach for automated and adaptive reconfiguration of multiple UAVs in a coordinated manner without the need for human intervention. This is suited to both military and civilian applications. Simulation results demonstrate the convergence and configuration of the UAVs with arbitrary initial locations and orientations. The performance has been tested for various number of UAVs and different mobility patterns of the ground target.