M.ENG in Robotics @ University of Maryland, College Park
Hello World! I am Moumita Paul, a full-time robot enthusiast, and a part-time wanderer. I have grown up seeing my father, experimenting with different types of electronic equipment in his maintenance lab in Indian Railways. Thus, engineering and I have coexisted for a long time.
The aspiration to forge a preferable life for every human being has formulated me to opt for pursuing research at Geometric Algorithms for Modeling, Motion, and Animation (GAMMA) Lab, UMD. I wish to dedicate the rest of my life to work in the development of robotics so that everyone will have a good quality of life.
My areas of focus are Computer Vision, Motion Planning, and Assistive Technology.
I am also a member of the Society of Women Engineers. On weekends, to blow off steam, either I paint, or travel to random new places. I did cliff jumping and someday you might see me bale out of an airplane too. And it might sound strange but I love long walks on snowy evenings.
University of Maryland
Masters of Engineering in Robotics; GPA: 3.75 / 4
1. Perception for Autonomous Robots.
2. Decision-Making for Robotics.
3. Advanced Techniques in Visual Learning & Recognition
4. AI Planning.
5. Planning for Autonomous Robots.
6. Control of Robotic Systems.
7. Introduction to Robot Modeling.
College Park, MD
Expected May, 2021
West Bengal University of Technology
B.Tech in Mechanical Engineering; Among Top 5%
- Basic Computation & Principles of Computer Programming.
- Numerical Methods.
- Basic Electrical & Electronics Engineering-I
- Basic Electrical & Electronics Engineering-II
- Engineering Mechanics.
- Dynamics of Machines.
- Design of Machine Elements.
- Machine Design.
- Industrial Instrumentation.
Relevant Work Expereince
Geometric Algorithms for Modeling, Motion, and Animation (GAMMA) Lab
May 2020- Present
Computer Vision RESEARCH ASSISTANT
Social Distancing Robot: Working under supervision of Dr. Dinesh Manocha on a social distancing robot that focuses on detecting and displaying the present 6 feet social distancing norm in crowded public places. This project also focuses on detecting elevated temperatures in subjects using thermal camera.
Endoenergy Systems Ltd
Jan 2019 – Apr 2019
Robotics RESEARCH INTERN
Hip and Knee Exoskeleton: Designed and controlled a lower-body exoskeleton for hip and knee joint to provide support during walking and Sit to Stand motion. Real-time Human motion data was analyzed and simulated using OpenSim to calculate the force and torque of each joint of the lower body.
Indian Institute of Technology, Gandhinagar
Jan 2018 – March 2018
Knee Sound Sensing Device: The goal of the research was to construct a portable wearable device for sensing acoustical emissions from the knee and characterization of that sound to distinguish between healthy and deleterious knee. This work focuses on the early detection of the ruinous effect of osteoarthritis or any acute injury.
Dogs Vs Cats Classification
The Dogs vs. Cats dataset is a standard computer vision dataset that involves classifying photos as either containing a dog or cat.
Although the problem sounds simple, it was only effectively addressed in the last few years using deep learning convolutional neural networks. While the dataset is effectively solved, it can be used as the basis for learning and practicing how to develop, evaluate, and use convolutional deep learning neural networks for image classification from scratch.
This includes how to develop a robust test harness for estimating the performance of the model, how to explore improvements to the model, and how to save the model and later load it to make predictions on new data.
Basic pipeline for the Project:
- Read the dataset into Google Colab.
- Create a deep learning model.
- Define Parameters for training.
- Train the model.
- Evaluate the model by analyzing training accuracy, validation accuracy, training loss, and validation loss curves.
- Optimize the model by reducing overfitting, training data augmentation, tweaking the architecture, etc.
Lane Detection For Autonomous vehicles
This project focused on doing simple Lane Detection to mimic Lane Departure Warning systems used in Self Driving Cars.
The task was to design an algorithm to detect lanes on the road, as well as estimate the road curvature to predict car turns.
Basic pipeline for the project.
- Unwarp an image and colorspace
- Compute Hough lines
- Histogram Analysis
- Sliding window polyfit
- Turn Prediction.
Visual Odometry is a method of finding a robot/camera pose i.e translation and orientation of the robot/camera with respect to the world frame using camera data.
The study of sequential images and their corresponding points helps in calculating the displacement of the robot/camera.
The consecutive pairs of images were used to estimate the camera pose attached to the car in the data set. The set of images with the camera calibration matrix are the given parameters for this project and the output is the trajectory plot of the car/camera in the x and z-axis. The output is also compared with output using pre-defined Opencv functions. The basic concepts used for Visual Odometry can be compared to the results of SLAM which is an important topic in Image Processing and Perception.
Basic pipeline for the project.
- Feature Matching.
- RANSAC based Outlier Rejection.
- Estimation of Fundamental Matrix.
- Estimate Essential Matrix based on Epipolar geometry.
- Estimate and refine Camera Pose.
- Check Triangulation and Disambiguate Camera poses.
- Transformation of Camera pose.
Underwater buoy Detection
A buoy can be anchored to the bottom for designating moorings, navigable channels, or obstructions in a water body.
A buoy is a distinctly shaped, colored floating device that has many practical purposes. It can be anchored to the bottom for designating moorings, navigable channels, or obstructions in a water body. Buoys can also be useful for underwater markings for navigation. For this project, the given underwater video sequence shows three different colour buoys such as orange, yellow, and green.
Since the buoys are distinctly colored and shaped, one approach could have been segmenting them using the color technique but for the given environment where there are noise and varying light intensities, this approach will not be optimal. So the Gaussian Mixture Model is being used to detect each buoy of the video.
Basic pipeline for the project.
- Data Preparation
- Color Segmentation using 1 D Gaussian
- Gaussian Mixture Model (GMM)
- Expectation Maximization
- Buoy Detection
The field of Computer vision has diverse applications in today’s world. When used in conjunction with machine learning, computer vision is used to provide machines and computers human-like intelligence.