Abhinav Kumar

Abhinav Kumar, Garrick Brazil, Enrique Corona et al.

Modern neural networks use building blocks such as convolutions that are equivariant to arbitrary 2D translations. However, these vanilla blocks are not equivariant to arbitrary 3D translations in the projective manifold. Even then, all monocular 3D detectors use vanilla blocks to obtain the 3D coordinates, a task for which the vanilla blocks are not designed for. This paper takes the first step towards convolutions equivariant to arbitrary 3D translations in the projective manifold. Since the depth is the hardest to estimate for monocular detection, this paper proposes Depth EquiVarIAnt NeTwork (DEVIANT) built with existing scale equivariant steerable blocks. As a result, DEVIANT is equivariant to the depth translations in the projective manifold whereas vanilla networks are not. The additional depth equivariance forces the DEVIANT to learn consistent depth estimates, and therefore, DEVIANT achieves state-of-the-art monocular 3D detection results on KITTI and Waymo datasets in the image-only category and performs competitively to methods using extra information. Moreover, DEVIANT works better than vanilla networks in cross-dataset evaluation. Code and models at https://github.com/abhi1kumar/DEVIANT

Abhinav Kumar, Chenhao Tan, Amit Sharma

Neural network models trained on text data have been found to encode undesirable linguistic or sensitive concepts in their representation. Removing such concepts is non-trivial because of a complex relationship between the concept, text input, and the learnt representation. Recent work has proposed post-hoc and adversarial methods to remove such unwanted concepts from a model's representation. Through an extensive theoretical and empirical analysis, we show that these methods can be counter-productive: they are unable to remove the concepts entirely, and in the worst case may end up destroying all task-relevant features. The reason is the methods' reliance on a probing classifier as a proxy for the concept. Even under the most favorable conditions for learning a probing classifier when a concept's relevant features in representation space alone can provide 100% accuracy, we prove that a probing classifier is likely to use non-concept features and thus post-hoc or adversarial methods will fail to remove the concept correctly. These theoretical implications are confirmed by experiments on models trained on synthetic, Multi-NLI, and Twitter datasets. For sensitive applications of concept removal such as fairness, we recommend caution against using these methods and propose a spuriousness metric to gauge the quality of the final classifier.

Abhinav Kumar, Barbara Di Eugenio, Abari Bhattacharya et al.

Reference resolution, which aims to identify entities being referred to by a speaker, is more complex in real world settings: new referents may be created by processes the agents engage in and/or be salient only because they belong to the shared physical setting. Our focus is on resolving references to visualizations on a large screen display in multimodal dialogue; crucially, reference resolution is directly involved in the process of creating new visualizations. We describe our annotations for user references to visualizations appearing on a large screen via language and hand gesture and also new entity establishment, which results from executing the user request to create a new visualization. We also describe our reference resolution pipeline which relies on an information-state architecture to maintain dialogue context. We report results on detecting and resolving references, effectiveness of contextual information on the model, and under-specified requests for creating visualizations. We also experiment with conventional CRF and deep learning / transformer models (BiLSTM-CRF and BERT-CRF) for tagging references in user utterance text. Our results show that transfer learning significantly boost performance of the deep learning methods, although CRF still out-performs them, suggesting that conventional methods may generalize better for low resource data.

Aniket Vashishtha, Abbavaram Gowtham Reddy, Abhinav Kumar et al.

Large Language Models (LLMs) have been used as experts to infer causal graphs, often by repeatedly applying a pairwise prompt that asks about the causal relationship of each variable pair. However, such experts, including human domain experts, cannot distinguish between direct and indirect effects given a pairwise prompt. Therefore, instead of the graph, we propose that causal order be used as a more stable output interface for utilizing expert knowledge. Even when querying a perfect expert with a pairwise prompt, we show that the inferred graph can have significant errors whereas the causal order is always correct. In practice, however, LLMs are imperfect experts and we find that pairwise prompts lead to multiple cycles. Hence, we propose the triplet method, a novel querying strategy that introduces an auxiliary variable for every variable pair and instructs the LLM to avoid cycles within this triplet. It then uses a voting-based ensemble method that results in higher accuracy and fewer cycles while ensuring cost efficiency. Across multiple real-world graphs, such a triplet-based method yields a more accurate order than the pairwise prompt, using both LLMs and human annotators. The triplet method enhances robustness by repeatedly querying an expert with different auxiliary variables, enabling smaller models like Phi-3 and Llama-3 8B Instruct to surpass GPT-4 with pairwise prompting. For practical usage, we show how the expert-provided causal order from the triplet method can be used to reduce error in downstream graph discovery and effect inference tasks.

Aniket Vashishtha, Abhinav Kumar, Atharva Pandey et al.

For text-based AI systems to interact in the real world, causal reasoning is an essential skill. Since active interventions are costly, we study to what extent a system can learn causal reasoning from symbolic demonstrations of causal axioms. Specifically, we present an axiomatic training method where the system learns from multiple demonstrations of a causal axiom (or rule), rather than incorporating the axiom as an inductive bias or inferring it from data values. A key question is whether the system would learn to generalize from the axiom demonstrations to more complex scenarios. Our results, based on applying axiomatic training to learn the transitivity axiom and d-separation rule, indicate that such generalization is possible. To avoid data contamination issues, we start with a 67 million parameter transformer model and train it from scratch. On both tasks, we find that a model trained on linear causal chains (along with some noisy variations) can generalize well to complex graphs, including longer causal chains, causal chains with reversed order, and graphs with branching.To handle diverse text inputs, the same method is extended to finetune language models. Finetuning Llama-3-8B-Instruct model on our axiomatic data leads to significant gains on causal benchmarks such as Corr2Cause and CLEAR, in some cases providing state-of-the-art performance surpassing GPT-4.

Yunfei Long, Abhinav Kumar, Xiaoming Liu et al.

Radar hits reflect from points on both the boundary and internal to object outlines. This results in a complex distribution of radar hits that depends on factors including object category, size, and orientation. Current radar-camera fusion methods implicitly account for this with a black-box neural network. In this paper, we explicitly utilize a radar hit distribution model to assist fusion. First, we build a model to predict radar hit distributions conditioned on object properties obtained from a monocular detector. Second, we use the predicted distribution as a kernel to match actual measured radar points in the neighborhood of the monocular detections, generating matching scores at nearby positions. Finally, a fusion stage combines context with the kernel detector to refine the matching scores. Our method achieves the state-of-the-art radar-camera detection performance on nuScenes. Our source code is available at https://github.com/longyunf/riccardo.

Feras A. Batarseh, Abhinav Kumar, Sam Eisenberg

Open data for all New Yorkers is the tagline on New York City's open data website. Open government is being promoted at most countries of the western world. Government transparency levels are being measured by the amount of data they share through their online public repositories. Additionally, open source software is promoted at governments, academia, and the industry. This is the new digital story of this century, and the new testament between the Gods of technology and there users. Data and software openness will redefine the path forward and aim to rekindle our collective intelligence. Data and software openness can redefine Data Democracy and be the catalyst for its progress. This chapter provides a historical insight into data and software openness, the beginnings, the heroes, prospects for the future, and all things we cannot afford to negotiate or lose.

Abhinav Kumar, Garrick Brazil, Xiaoming Liu

Modern 3D object detectors have immensely benefited from the end-to-end learning idea. However, most of them use a post-processing algorithm called Non-Maximal Suppression (NMS) only during inference. While there were attempts to include NMS in the training pipeline for tasks such as 2D object detection, they have been less widely adopted due to a non-mathematical expression of the NMS. In this paper, we present and integrate GrooMeD-NMS -- a novel Grouped Mathematically Differentiable NMS for monocular 3D object detection, such that the network is trained end-to-end with a loss on the boxes after NMS. We first formulate NMS as a matrix operation and then group and mask the boxes in an unsupervised manner to obtain a simple closed-form expression of the NMS. GrooMeD-NMS addresses the mismatch between training and inference pipelines and, therefore, forces the network to select the best 3D box in a differentiable manner. As a result, GrooMeD-NMS achieves state-of-the-art monocular 3D object detection results on the KITTI benchmark dataset performing comparably to monocular video-based methods. Code and models at https://github.com/abhi1kumar/groomed_nms

Thiago Serra, Xin Yu, Abhinav Kumar et al.

We can compress a rectifier network while exactly preserving its underlying functionality with respect to a given input domain if some of its neurons are stable. However, current approaches to determine the stability of neurons with Rectified Linear Unit (ReLU) activations require solving or finding a good approximation to multiple discrete optimization problems. In this work, we introduce an algorithm based on solving a single optimization problem to identify all stable neurons. Our approach is on median 183 times faster than the state-of-art method on CIFAR-10, which allows us to explore exact compression on deeper (5 x 100) and wider (2 x 800) networks within minutes. For classifiers trained under an amount of L1 regularization that does not worsen accuracy, we can remove up to 56% of the connections on the CIFAR-10 dataset. The code is available at the following link, https://github.com/yuxwind/ExactCompression.

Shengjie Zhu, Girish Chandar Ganesan, Abhinav Kumar et al.

3D sensing is a fundamental task for Autonomous Vehicles. Its deployment often relies on aligned RGB cameras and LiDAR. Despite meticulous synchronization and calibration, systematic misalignment persists in LiDAR projected depthmap. This is due to the physical baseline distance between the two sensors. The artifact is often reflected as background LiDAR incorrectly projected onto the foreground, such as cars and pedestrians. The KITTI dataset uses stereo cameras as a heuristic solution to remove artifacts. However most AV datasets, including nuScenes, Waymo, and DDAD, lack stereo images, making the KITTI solution inapplicable. We propose RePLAy, a parameter-free analytical solution to remove the projective artifacts. We construct a binocular vision system between a hypothesized virtual LiDAR camera and the RGB camera. We then remove the projective artifacts by determining the epipolar occlusion with the proposed analytical solution. We show unanimous improvement in the State-of-The-Art (SoTA) monocular depth estimators and 3D object detectors with the artifacts-free depthmaps.

Yuliang Guo, Abhinav Kumar, Cheng Zhao et al.

Monocular 3D reconstruction for categorical objects heavily relies on accurately perceiving each object's pose. While gradient-based optimization in a NeRF framework updates the initial pose, this paper highlights that scale-depth ambiguity in monocular object reconstruction causes failures when the initial pose deviates moderately from the true pose. Consequently, existing methods often depend on a third-party 3D object to provide an initial object pose, leading to increased complexity and generalization issues. To address these challenges, we present SUP-NeRF, a Streamlined Unification of object Pose estimation and NeRF-based object reconstruction. SUP-NeRF decouples the object's dimension estimation and pose refinement to resolve the scale-depth ambiguity, and introduces a camera-invariant projected-box representation that generalizes cross different domains. While using a dedicated pose estimator that smoothly integrates into an object-centric NeRF, SUP-NeRF is free from external 3D detectors. SUP-NeRF achieves state-of-the-art results in both reconstruction and pose estimation tasks on the nuScenes dataset. Furthermore, SUP-NeRF exhibits exceptional cross-dataset generalization on the KITTI and Waymo datasets, surpassing prior methods with up to 50\% reduction in rotation and translation error.

Md. Nurul Absur, Abhinav Kumar, Swastik Brahma et al.

Multi-view 3D reconstruction applications are revolutionizing critical use cases that require rapid situational-awareness, such as emergency response, tactical scenarios, and public safety. In many cases, their near-real-time latency requirements and ad-hoc needs for compute resources necessitate adoption of `Just-in-time' edge environments where the system is set up on the fly to support the applications during the mission lifetime. However, reliability issues can arise from the inherent dynamism and operational adversities of such edge environments, resulting in spatiotemporally correlated disruptions that impact the camera operations, which can lead to sustained degradation of reconstruction quality. In this paper, we propose a novel portfolio theory inspired edge resource management strategy for reliable multi-view 3D reconstruction against possible system disruptions. Our proposed methodology can guarantee reconstruction quality satisfaction even when the cameras are prone to spatiotemporally correlated disruptions. The portfolio theoretic optimization problem is solved using a genetic algorithm that converges quickly for realistic system settings. Using publicly available and customized 3D datasets, we demonstrate the proposed camera selection strategy's benefits in guaranteeing reliable 3D reconstruction against traditional baseline strategies, under spatiotemporal disruptions.

Zhihao Zhang, Abhinav Kumar, Girish Chandar Ganesan et al.

Accurately predicting 3D attributes is crucial for monocular 3D object detection (Mono3D), with depth estimation posing the greatest challenge due to the inherent ambiguity in mapping 2D images to 3D space. While existing methods leverage multiple depth cues (e.g., estimating depth uncertainty, modeling depth error) to improve depth accuracy, they overlook that accurate depth prediction requires conditioning on other 3D attributes, as these attributes are intrinsically inter-correlated through the 3D to 2D projection, which ultimately limits overall accuracy and stability. Inspired by Chain-of-Thought (CoT) in large language models (LLMs), this paper proposes MonoCoP, which leverages a Chain-of-Prediction (CoP) to predict attributes sequentially and conditionally via three key designs. First, it employs a lightweight AttributeNet (AN) for each 3D attribute to learn attribute-specific features. Next, MonoCoP constructs an explicit chain to propagate these learned features from one attribute to the next. Finally, MonoCoP uses a residual connection to aggregate features for each attribute along the chain, ensuring that later attribute predictions are conditioned on all previously processed attributes without forgetting the features of earlier ones. Experimental results show that our MonoCoP achieves state-of-the-art (SoTA) performance on the KITTI leaderboard without requiring additional data and further surpasses existing methods on the Waymo and nuScenes frontal datasets.

Ashwin Kalyan, Abhinav Kumar, Arjun Chandrasekaran et al.

Many real-world problems require the combined application of multiple reasoning abilities employing suitable abstractions, commonsense knowledge, and creative synthesis of problem-solving strategies. To help advance AI systems towards such capabilities, we propose a new reasoning challenge, namely Fermi Problems (FPs), which are questions whose answers can only be approximately estimated because their precise computation is either impractical or impossible. For example, "How much would the sea level rise if all ice in the world melted?" FPs are commonly used in quizzes and interviews to bring out and evaluate the creative reasoning abilities of humans. To do the same for AI systems, we present two datasets: 1) A collection of 1k real-world FPs sourced from quizzes and olympiads; and 2) a bank of 10k synthetic FPs of intermediate complexity to serve as a sandbox for the harder real-world challenge. In addition to question answer pairs, the datasets contain detailed solutions in the form of an executable program and supporting facts, helping in supervision and evaluation of intermediate steps. We demonstrate that even extensively fine-tuned large scale language models perform poorly on these datasets, on average making estimates that are off by two orders of magnitude. Our contribution is thus the crystallization of several unsolved AI problems into a single, new challenge that we hope will spur further advances in building systems that can reason.

Feras A. Batarseh, Rasika Mohod, Abhinav Kumar et al.

The field of artificial intelligence (AI) is witnessing a recent upsurge in research, tools development, and deployment of applications. Multiple software companies are shifting their focus to developing intelligent systems; and many others are deploying AI paradigms to their existing processes. In parallel, the academic research community is injecting AI paradigms to provide solutions to traditional engineering problems. Similarly, AI has evidently been proved useful to software engineering (SE). When one observes the SE phases (requirements, design, development, testing, release, and maintenance), it becomes clear that multiple AI paradigms (such as neural networks, machine learning, knowledge-based systems, natural language processing) could be applied to improve the process and eliminate many of the major challenges that the SE field has been facing. This survey chapter is a review of the most commonplace methods of AI applied to SE. The review covers methods between years 1975-2017, for the requirements phase, 46 major AI-driven methods are found, 19 for design, 15 for development, 68 for testing, and 15 for release and maintenance. Furthermore, the purpose of this chapter is threefold; firstly, to answer the following questions: is there sufficient intelligence in the SE lifecycle? What does applying AI to SE entail? Secondly, to measure, formulize, and evaluate the overlap of SE phases and AI disciplines. Lastly, this chapter aims to provide serious questions to challenging the current conventional wisdom (i.e., status quo) of the state-of-the-art, craft a call for action, and to redefine the path forward.

Raj Jagtap, Abhinav Kumar, Rahul Goel et al.

Millions of people use platforms such as YouTube, Facebook, Twitter, and other mass media. Due to the accessibility of these platforms, they are often used to establish a narrative, conduct propaganda, and disseminate misinformation. This work proposes an approach that uses state-of-the-art NLP techniques to extract features from video captions (subtitles). To evaluate our approach, we utilize a publicly accessible and labeled dataset for classifying videos as misinformation or not. The motivation behind exploring video captions stems from our analysis of videos metadata. Attributes such as the number of views, likes, dislikes, and comments are ineffective as videos are hard to differentiate using this information. Using caption dataset, the proposed models can classify videos among three classes (Misinformation, Debunking Misinformation, and Neutral) with 0.85 to 0.90 F1-score. To emphasize the relevance of the misinformation class, we re-formulate our classification problem as a two-class classification - Misinformation vs. others (Debunking Misinformation and Neutral). In our experiments, the proposed models can classify videos with 0.92 to 0.95 F1-score and 0.78 to 0.90 AUC ROC.

Abhinav Kumar, Tim K. Marks, Wenxuan Mou et al.

Modern face alignment methods have become quite accurate at predicting the locations of facial landmarks, but they do not typically estimate the uncertainty of their predicted locations nor predict whether landmarks are visible. In this paper, we present a novel framework for jointly predicting landmark locations, associated uncertainties of these predicted locations, and landmark visibilities. We model these as mixed random variables and estimate them using a deep network trained with our proposed Location, Uncertainty, and Visibility Likelihood (LUVLi) loss. In addition, we release an entirely new labeling of a large face alignment dataset with over 19,000 face images in a full range of head poses. Each face is manually labeled with the ground-truth locations of 68 landmarks, with the additional information of whether each landmark is unoccluded, self-occluded (due to extreme head poses), or externally occluded. Not only does our joint estimation yield accurate estimates of the uncertainty of predicted landmark locations, but it also yields state-of-the-art estimates for the landmark locations themselves on multiple standard face alignment datasets. Our method's estimates of the uncertainty of predicted landmark locations could be used to automatically identify input images on which face alignment fails, which can be critical for downstream tasks.

Thiago Serra, Abhinav Kumar, Srikumar Ramalingam

Deep neural networks have been successful in many predictive modeling tasks, such as image and language recognition, where large neural networks are often used to obtain good accuracy. Consequently, it is challenging to deploy these networks under limited computational resources, such as in mobile devices. In this work, we introduce an algorithm that removes units and layers of a neural network while not changing the output that is produced, which thus implies a lossless compression. This algorithm, which we denote as LEO (Lossless Expressiveness Optimization), relies on Mixed-Integer Linear Programming (MILP) to identify Rectified Linear Units (ReLUs) with linear behavior over the input domain. By using L1 regularization to induce such behavior, we can benefit from training over a larger architecture than we would later use in the environment where the trained neural network is deployed.

Abhinav Kumar, Thiago Serra, Srikumar Ramalingam

Two networks are equivalent if they produce the same output for any given input. In this paper, we study the possibility of transforming a deep neural network to another network with a different number of units or layers, which can be either equivalent, a local exact approximation, or a global linear approximation of the original network. On the practical side, we show that certain rectified linear units (ReLUs) can be safely removed from a network if they are always active or inactive for any valid input. If we only need an equivalent network for a smaller domain, then more units can be removed and some layers collapsed. On the theoretical side, we constructively show that for any feed-forward ReLU network, there exists a global linear approximation to a 2-hidden-layer shallow network with a fixed number of units. This result is a balance between the increasing number of units for arbitrary approximation with a single layer and the known upper bound of $\lceil log(n_0+1)\rceil +1$ layers for exact representation, where $n_0$ is the input dimension. While the transformed network may require an exponential number of units to capture the activation patterns of the original network, we show that it can be made substantially smaller by only accounting for the patterns that define linear regions. Based on experiments with ReLU networks on the MNIST dataset, we found that $l_1$-regularization and adversarial training reduces the number of linear regions significantly as the number of stable units increases due to weight sparsity. Therefore, we can also intentionally train ReLU networks to allow for effective loss-less compression and approximation.

Mayank Gupta, Abhinav Kumar, Sriganesh Madhvanath

We describe a novel method of generating high-resolution real-world images of text where the style and textual content of the images are described parametrically. Our method combines text to image retrieval techniques with progressive growing of Generative Adversarial Networks (PGGANs) to achieve conditional generation of photo-realistic images that reflect specific styles, as well as artifacts seen in real-world images. We demonstrate our method in the context of automotive license plates. We assess the impact of varying the number of training images of each style on the fidelity of the generated style, and demonstrate the quality of the generated images using license plate recognition systems.

Nagabhushan Eswara, S Ashique, Anand Panchbhai et al.

HTTP based adaptive video streaming has become a popular choice of streaming due to the reliable transmission and the flexibility offered to adapt to varying network conditions. However, due to rate adaptation in adaptive streaming, the quality of the videos at the client keeps varying with time depending on the end-to-end network conditions. Further, varying network conditions can lead to the video client running out of playback content resulting in rebuffering events. These factors affect the user satisfaction and cause degradation of the user quality of experience (QoE). It is important to quantify the perceptual QoE of the streaming video users and monitor the same in a continuous manner so that the QoE degradation can be minimized. However, the continuous evaluation of QoE is challenging as it is determined by complex dynamic interactions among the QoE influencing factors. Towards this end, we present LSTM-QoE, a recurrent neural network based QoE prediction model using a Long Short-Term Memory (LSTM) network. The LSTM-QoE is a network of cascaded LSTM blocks to capture the nonlinearities and the complex temporal dependencies involved in the time varying QoE. Based on an evaluation over several publicly available continuous QoE databases, we demonstrate that the LSTM-QoE has the capability to model the QoE dynamics effectively. We compare the proposed model with the state-of-the-art QoE prediction models and show that it provides superior performance across these databases. Further, we discuss the state space perspective for the LSTM-QoE and show the efficacy of the state space modeling approaches for QoE prediction.

Abhinav Kumar, Shantanu Gupta, Vladimir Kozitsky et al.

The problem of vehicle licence plate re-identification is generally considered as a one-shot image retrieval problem. The objective of this task is to learn a feature representation (called a "signature") for licence plates. Incoming licence plate images are converted to signatures and matched to a previously collected template database through a distance measure. Then, the input image is recognized as the template whose signature is "nearest" to the input signature. The template database is restricted to contain only a single signature per unique licence plate for our problem. We measure the performance of deep convolutional net-based features adapted from face recognition on this task. In addition, we also test a hybrid approach combining the Fisher vector with a neural network-based embedding called "f2nn" trained with the Triplet loss function. We find that the hybrid approach performs comparably while providing computational benefits. The signature generated by the hybrid approach also shows higher generalizability to datasets more dissimilar to the training corpus.