### Shubham Shrivastava

Technical Expert and Lead @Ford Autonomy3D Perception

### Education

Stanford University

GPA: 4+/4.0

Courses in Meta-Learning, Multi-Task Learning, Deep Generative Models, Natural Language Processing, Computer Vision, and 3D Reconstruction

August 2020 - Present

The University of Texas at Arlington

Master of Science in Electrical Engineering

GPA: 4.0/4.0

August 2014 - August 2016

Visvesvaraya Technological University

Bachelor of Engineering in Electronics and Communication Engineering

GPA: 4.0/4.0, First Class with Distinction, Aggregate Percentage: 86%

August 2014 - August 2016

### Work Experience

Ford Greenfield Labs, Palo Alto, CA

Technical Expert and Lead - 3D Perception

September 2022 - Present

Leading vision-centric 3D perception workstream at Ford Autonomy towards building magical experiences; including (1) end-to-end 3D Perception stack development for future Ford L3 vehicles on road, (2) machine-learning training framework for all ML tasks within Ford.

Research Scientist  - Machine Learning and Computer Vision

September 2019 - September 2022

My research includes computer vision and advanced machine learning methods including convolutional neural networks, generative adversarial networks, variational autoencoders, and 3D perception; with significant emphasis on object detection, semantics learning, 3D scene understanding, multi-view geometry, and visual odometry.

[Topics of Research]

Renesas Electronics America, Inc.

Applications Engineer, Perception R&D - ADAS and Autonomous Driving

March 2017 - September 2019

I worked as part of a very small team towards building the ADAS & Autonomous Driving perception reference platform “Perception Quick Start” which includes end-to-end solutions for camera and LiDAR based road feature and object detection.

Changan US Research & Development Center, Inc.

Intelligent Vehicle Engineer (Connected Autonomous Vehicle Research Group)

August 2016 - March 2017

Worked within Changan's Connected and Autonomous Vehicle research team to design and develop various vehicle safety models for 360-degrees target classification and warnings notifications with and without line-of-sight requirements.

DELPHI (now known as APTIV)

Embedded Software Engineer (Intern)

May 2016 - August 2016

Worked with application teams, forward systems algorithm group, and controller design groups to define the functionality, develop algorithms, and implement them in accordance with the V-Model Software Development Life Cycle.

BlackBerry QNX

Software Development Intern (Board Support Package)

January 2016 - May 2016

Developed the BSP (Board Support Package) for custom hardware with i.MX6 Solo Processor and several peripherals. Worked towards low-level board bring-ups and provided support for the following peripherals.

The University of Texas at Arlington Research Institute

Research Intern

August 2015 - December 2015

Designed and Developed the control GUI for a prosthetic system used to help rehabilitate post-stroke patients. It used an Arduino controller for adaptive adjustment of the air bubble pressure at the desired psi value for various points on the leg. Also designed a GUI which allows the user to enter the desired psi values for each air bubble and simultaneously measure current bubble pressure and display it on the GUI in real-time.

Indian Institute of Science (IISc)

Trainee Engineer

January 2014 - May 2014

Designed and developed a 2 Dimensional plotter (Smart XY Plotter) at the Mechatronics Lab, IISc, capable of plotting any 2D image using a pen, which was controlled by means of a Solenoid and two stepper motors (responsible for x, y, and z directional movement).

### Papers and Publications

ICRA 2022

Propagating State Uncertainty Through Trajectory Forecasting

Boris Ivanovic, Yifeng Lin, Shubham Shrivastava, Punarjay Chakravarty, Marco Pavone
Uncertainty pervades through the modern robotic autonomy stack, with nearly every component (e.g., sensors, detection, classification, tracking, behavior prediction) producing continuous or discrete probabilistic distributions. Trajectory forecasting, in particular, is surrounded by uncertainty as its inputs are produced by (noisy) upstream perception and its outputs are predictions that are often probabilistic for use in downstream planning. However, most trajectory forecasting methods do not account for upstream uncertainty, instead taking only the most-likely values. As a result, perceptual uncertainties are not propagated through forecasting and predictions are frequently overconfident. To address this, we present a novel method for incorporating perceptual state uncertainty in trajectory forecasting, a key component of which is a new statistical distance-based loss function which encourages predicting uncertainties that better match upstream perception. We evaluate our approach both in illustrative simulations and on large-scale, real-world data, demonstrating its efficacy in propagating perceptual state uncertainty through prediction and producing more calibrated predictions.
BMVC 2022

Category-Level Pose Retrieval with Contrastive Features Learnt with Occlusion Augmentation

Georgios Kouros, Shubham Shrivastava, Cédric Picron, Sushruth Nagesh, Punarjay Chakravarty, Tinne Tuytelaars
Pose estimation is usually tackled as either a bin classification problem or as a regression problem. In both cases, the idea is to directly predict the pose of an object. This is a non-trivial task because of appearance variations of similar poses and similarities between different poses. Instead, we follow the key idea that it is easier to compare two poses than to estimate them. Render-and-compare approaches have been employed to that end, however, they tend to be unstable, computationally expensive, and slow for real-time applications. We propose doing category-level pose estimation by learning an alignment metric using a contrastive loss with a dynamic margin and a continuous pose-label space. For efficient inference, we use a simple real-time image retrieval scheme with a reference set of renderings projected to an embedding space. To achieve robustness to real-world conditions, we employ synthetic occlusions, bounding box perturbations, and appearance augmentations. Our approach achieves state-of-the-art performance on PASCAL3D and OccludedPASCAL3D, as well as high-quality results on KITTI3D.
IJCAI 2021 AI for Autonomous Driving Workshop
3D object detection and dense depth estimation are one of the most vital tasks in autonomous driving. Multiple sensor modalities can jointly attribute towards better robot perception, and to that end, we introduce a method for jointly training 3D object detection and monocular dense depth reconstruction neural networks. It takes as inputs, a LiDAR point-cloud, and a single RGB image during inference and produces object pose predictions as well as a densely reconstructed depth map. LiDAR point-cloud is converted into a set of voxels, and its features are extracted using 3D convolution layers, from which we regress object pose parameters. Corresponding RGB image features are extracted using another 2D convolutional neural network. We further use these combined features to predict a dense depth map. While our object detection is trained in a supervised manner, the depth prediction network is trained with both self-supervised and supervised loss functions. We also introduce a loss function, edge-preserving smooth loss, and show that this results in better depth estimation compared to the edge-aware smooth loss function, frequently used in depth prediction works.
IROS 2022 + RA-L

Improving Worst Case Visual Localization Coverage via Place-specific Sub-selection in Multi-camera Systems

Stephen Hausler, Ming Xu, Sourav Garg, Punarjay Chakravarty, Shubham Shrivastava, Ankit Vora, Michael Milford
6-DoF visual localization systems utilize principled approaches rooted in 3D geometry to perform accurate camera pose estimation of images to a map. Current techniques use hierarchical pipelines and learned 2D feature extractors to improve scalability and increase performance. However, despite gains in typical recall@0.25m type metrics, these systems still have limited utility for real-world applications like autonomous vehicles because of their worst' areas of performance - the locations where they provide insufficient recall at a certain required error tolerance. Here we investigate the utility of using place specific configurations', where a map is segmented into a number of places, each with its own configuration for modulating the pose estimation step, in this case selecting a camera within a multi-camera system. On the Ford AV benchmark dataset, we demonstrate substantially improved worst-case localization performance compared to using off-the-shelf pipelines - minimizing the percentage of the dataset which has low recall at a certain error tolerance, as well as improved overall localization performance. Our proposed approach is particularly applicable to the crowdsharing model of autonomous vehicle deployment, where a fleet of AVs are regularly traversing a known route.
CVPR 2020 Workshop

Deflating Dataset Bias Using Synthetic Data Augmentation

Nikita Jaipuria, Xianling Zhang, Rohan Bhasin, Mayar Arafa, Punarjay Chakravarty, Shubham Shrivastava, Sagar Manglani, Vidya N. Murali

Deep Learning has seen an unprecedented increase in vision applications since the publication of large-scale object recognition datasets and the introduction of scalable compute hardware. State-of-the-art methods for most vision tasks for Autonomous Vehicles (AVs) rely on supervised learning and often fail to generalize to domain shifts and/or outliers. Dataset diversity is thus key to successful real-world deployment. No matter how big the size of the dataset, capturing long tails of the distribution pertaining to task-specific environmental factors is impractical. The goal of this paper is to investigate the use of targeted synthetic data augmentation - combining the benefits of gaming engine simulations and sim2real style transfer techniques - for filling gaps in real datasets for vision tasks. Empirical studies on three different computer vision tasks of practical use to AVs - parking slot detection, lane detection, and monocular depth estimation - consistently show that having synthetic data in the training mix provides a significant boost in cross-dataset generalization performance as compared to training on real data only, for the same size of the training set.

We introduce a method for 3D object detection using a single monocular image. Starting from a synthetic dataset, we pre-train an RGB-to-Depth Auto-Encoder (AE). The embedding learnt from this AE is then used to train a 3D Object Detector (3DOD) CNN which is used to regress the parameters of 3D object poses after the encoder from the AE generates a latent embedding from the RGB image. We show that we can pre-train the AE using paired RGB and depth images from simulation data once and subsequently only train the 3DOD network using real data, comprising of RGB images and 3D object pose labels (without the requirement of dense depth). Our 3DOD network utilizes a particular `cubification' of 3D space around the camera, where each cuboid is tasked with predicting N object poses, along with their class and confidence values. The AE pre-training and this method of dividing the 3D space around the camera into cuboids give our method its name - CubifAE-3D. We demonstrate results for monocular 3D object detection in the Autonomous Vehicle (AV) use-case with the Virtual KITTI 2 and the KITTI datasets.

Training models that are robust to data domain shift have gained an increasing interest both in academia and industry. Question-Answering language models, being one of the typical problem in Natural Language Processing (NLP) research, has received much success with the advent of large transformer models. However, existing approaches mostly work under the assumption that data is drawn from same distribution during training and testing which is unrealistic and non-scalable in the wild.

In this paper, we explore adversarial training approach towards learning domain-invariant features so that language models can generalize well to out-of-domain datasets. We also inspect various other ways to boost our model performance including data augmentation by paraphrasing sentences, conditioning end of answer span prediction on the start word, and carefully designed annealing function. Our initial results shows that in combination with these methods, we are able to achieve 15.2% improvement in EM score and 5.6% boost in F1 score on out-of-domain validation datasets over the baseline. We also dissect our model outputs and visualize the model hidden-states by projecting them onto a lower-dimensional space, and discover that our specific adversarial training approach indeed encourages the model to learn domain invariant embedding and bring them closer in the multi-dimensional space.

PoseGen: Pose-Conditioned Image Generation

Shubham Shrivastava and Amir Ziai

Autonomous driving application requires detection and localization of objects within a few centimeters. Deep learning models that can achieve this level of accuracy require millions of training data samples. Manually annotating data at this scale is highly cumbersome and expensive. Furthermore, annotating object poses in 3D is not possible from just a single image without requiring either additional sensors or multi-frame non-linear optimization. In this work we present \textit{PoseGen}, a methodology for generating a realistic image of an object given a desired pose, appearance, and background. We do this in an unsupervised way by conditioning the generation process on various attributes. We also release a dataset of paired objects and silhouette masks TeslaPose dataset which we hope will help the research community further tackle this problem

Meta-Regularization by Enforcing Mutual-Exclusiveness

Shubham Shrivastava, Edwin Pan, and Pankaj Rajak

An A* Curriculum Approach to Reinforcement Learning for RGBD Indoor Robot Navigation

Kaushik Balakrishnan, Punarjay Chakravarty, Shubham Shrivastava
Training robots to navigate diverse environments is a challenging problem as it involves the confluence of several different perception tasks such as mapping and localization, followed by optimal path-planning and control. Recently released photo-realistic simulators such as Habitat allow for the training of networks that output control actions directly from perception: agents use Deep Reinforcement Learning (DRL) to regress directly from the camera image to a control output in an end-to-end fashion. This is data-inefficient and can take several days to train on a GPU. Our paper tries to overcome this problem by separating the training of the perception and control neural nets and increasing the path complexity gradually using a curriculum approach. Specifically, a pre-trained twin Variational AutoEncoder (VAE) is used to compress RGBD (RGB & depth) sensing from an environment into a latent embedding, which is then used to train a DRL-based control policy. A*, a traditional path-planner is used as a guide for the policy and the distance between start and target locations is incrementally increased along the A* route, as training progresses. We demonstrate the efficacy of the proposed approach, both in terms of increased performance and decreased training times for the PointNav task in the Habitat simulation environment. This strategy of improving the training of direct-perception based DRL navigation policies is expected to hasten the deployment of robots of particular interest to industry such as co-bots on the factory floor and last-mile delivery robots.
Computer Vision Conference (CVC) 2019 Conference Paper
In recent years, autonomous driving has inexorably progressed from the domain of science fiction to reality. For a self-driving car, it is of utmost importance that it knows its surroundings. Several sensors like RADARs, LiDARs, and Cameras have been primarily used to sense the environment and make a judgment on the next course of action. Object detection is of a great significance in Autonomous Driving wherein the self-driving car needs to identify the objects around it and must take necessary actions to avoid a collision. Several perception-based methods like classical Computer Vision techniques and Convolutional Neural Networks (CNN) exist today which detects and classifies an object. This paper discusses an object detection technique based on Stereo Vision. One challenge in this process though is to eliminate regions of the image which are insignificant for the detection, like unoccupied road and buildings far ahead. This paper proposes a method to first get rid of such regions using V-Disparity and then detect objects using 3D density-based clustering. Results given in this paper show that the proposed system can detect objects on the road very accurately and robustly.

S-BEV: Semantic Birds-Eye View Representation for Weather and Lighting Invariant 3-DoF Localization

Mokshith Voodarla, Shubham Shrivastava, Sagar Manglani, Ankit Vora, Siddharth Agarwal, Punarjay Chakravarty
We describe a light-weight, weather and lighting invariant, Semantic Bird's Eye View (S-BEV) signature for vision-based vehicle re-localization. A topological map of S-BEV signatures is created during the first traversal of the route, which are used for coarse localization in subsequent route traversal. A fine-grained localizer is then trained to output the global 3-DoF pose of the vehicle using its S-BEV and its coarse localization. We conduct experiments on vKITTI2 virtual dataset and show the potential of the S-BEV to be robust to weather and lighting. We also demonstrate results with 2 vehicles on a 22 km long highway route in the Ford AV dataset.

Sim2Real for Self-Supervised Monocular Depth and Segmentation

Nithin Raghavan, Punarjay Chakravarty, Shubham Shrivastava
Image-based learning methods for autonomous vehicle perception tasks require large quantities of labelled, real data in order to properly train without overfitting, which can often be incredibly costly. While leveraging the power of simulated data can potentially aid in mitigating these costs, networks trained in the simulation domain usually fail to perform adequately when applied to images in the real domain. Recent advances in domain adaptation have indicated that a shared latent space assumption can help to bridge the gap between the simulation and real domains, allowing the transference of the predictive capabilities of a network from the simulation domain to the real domain. We demonstrate that a twin VAE-based architecture with a shared latent space and auxiliary decoders is able to bridge the sim2real gap without requiring any paired, ground-truth data in the real domain. Using only paired, ground-truth data in the simulation domain, this architecture has the potential to generate perception tasks such as depth and segmentation maps. We compare this method to networks trained in a supervised manner to indicate the merit of these results.
Chapter in the Book: Connected Vehicles
National Highway Traffic Safety Administration (NTHSA) has been interested in vehicle-to-vehicle (V2V) communication as the next step in addressing grooving rates of fatalities from vehicle related crashes. Today’s crash avoidance technologies depend on on-board sensors like camera and radar to provide awareness input to the safety applications. These applications warn the driver of imminent danger or sometimes even act on the driver’s behalf. However, even technologies like those cannot “predict” a crash that might happen because of a vehicle which is not very close or not in the line of sight to the host vehicle. A technology that can “see” through another vehicle or obstacles like buildings and predict a danger can fill these gaps and reduce crashes drastically. V2V communications can provide vehicles the ability to talk to each other and therefore see around corners and through the obstacles over a longer distance compared to the current on-board sensors. It is estimated that V2X communications address up to 80% of the unimpaired crashes [1]. By means of Notice of Proposed Rulemaking (NPRM), NHTSA is working towards standardization of V2V communications and potentially mandating the broadcast of vehicle data (e.g. GPS coordinates, speed, acceleration) over DSRC through V2V.