Working with Qualcomm Automotive team on ADAS AI model performance and deployment. Responsibilities include model onboarding, model performance evaluation and improvement, and application development for deployment.
To enhance perception in autonomous vehicles (AVs), recent efforts are concentrating on 3D object detectors, which deliver more comprehensive predictions than traditional 2D object detectors, at the cost of increased memory footprint and computational resource usage. We present a novel framework called UPAQ, which leverages semi-structured pattern pruning and quantization to improve the efficiency of LiDAR point-cloud and camera-based 3D object detectors on resource-constrained embedded AV platforms. Experimental results on the Jetson Orin Nano embedded platform indicate that UPAQ achieves up to 5.62x and 5.13x model compression rates, up to 1.97x and 1.86x boost in inference speed, and up to 2.07x and 1.87x reduction in energy consumption compared to state-of-the-art model compression frameworks, on the Pointpillar and SMOKE models respectively.
Recent advances in machine learning (ML) have spotlighted the pressing need for computing architectures that bridge the gap between memory bandwidth and processing power. The advent of deep neural networks has pushed traditional Von Neumann architectures to their limits due to the high latency and energy consumption costs associated with data movement between the processor and memory for these workloads. One of the solutions to overcome this bottleneck is to perform computation within the main memory through processing-in-memory (PIM), thereby limiting data movement and the costs associated with it. However, dynamic random-access memory-based PIM struggles to achieve high throughput and energy efficiency due to internal data movement bottlenecks and the need for frequent refresh operations. In this work, we introduce OPIMA, a PIM-based ML accelerator, architected within an optical main memory. OPIMA has been designed to leverage the inherent massive parallelism within main memory while performing high-speed, low-energy optical computation to accelerate ML models based on convolutional neural networks. We present a comprehensive analysis of OPIMA to guide design choices and operational mechanisms. In addition, we evaluate the performance and energy consumption of OPIMA, comparing it with conventional electronic computing systems and emerging photonic PIM architectures. The experimental results show that OPIMA can achieve 2.98× higher throughput and 137× better energy efficiency than the best known prior work.
Object detectors used in autonomous vehicles can have high memory and computational overheads. In this paper, we introduce a novel semi-structured pruning framework called R-TOSS that overcomes the shortcomings of state-of-the-art model pruning techniques. Experimental results on the JetsonTX2 platform show that R-TOSS has a compression rate of 4.4× on the YOLOv5 object detector with a 2.15× speedup in inference time and 57.01% decrease in energy usage. R-TOSS also enables 2.89× compression on RetinaNet with a 1.86× speedup in inference time and 56.31% decrease in energy usage. We also demonstrate significant improvements compared to various state-of-the-art pruning techniques.
GTA for ECE 452 Computer Architecture and Organization taught by Dr Sudeep Pasricha consisting of 54 students. This role includes preparing assignments, resolving doubts and queries of students, and grading.
Object detection is a computer vision task that has become an integral part of many consumer applications today such as surveillance and security systems, mobile text recognition, and diagnosing diseases from MRI/CT scans. Object detection is also one of the critical components to support autonomous driving. Autonomous vehicles rely on the perception of their surroundings to ensure safe and robust driving performance. This perception system uses object detection algorithms to accurately determine objects such as pedestrians, vehicles, traffic signs, and barriers in the vehicle's vicinity. Deep learning-based object detectors play a vital role in finding and localizing these objects in real-time. This article discusses the state-of-the-art in object detectors and open challenges for their integration into autonomous vehicles.
Optimizing Deep Learning models using Pruning, quantization, knowledge distillation, encoding and decoding. Deployed on Jetson Tx2, Drive Px2, AWS sagemaker, AWS EC2, AWS Lambda, PaperSpace platforms. Technologies used - Keras, TensorFlow, CUDA, and cluster environments.
The Colorado State University Vehicle Innovation Team has continuously worked with AVTC’s since 1988, have has been successful in getting over 200 students to graduate from multiple departments. The CSU Vehicle Innovation Team this year hopes to follow the same morality as last year, by defining a solution to sustainable problems without compromising the ideals and expectations of the US consumer.
Pursuing my Master of Science in Electrical Engineering specializing in Embedded systems and Automation.
Cambionix Innovations provides various solutions in industrial and home automation.The company is built on the strong foundation of a synergetic skilled team. Our expertise includes specialists in robotics, automation, data mining, artificial intelligence, embedded systems, IoT & energy sector.
Our Vision is to provide Low cost reliable solutions for small and medium scale industries and increase there productivity.
I as a part of this journey, redefining the work process in several industries and helped them increase productivity and reduce operating cost
Aurora is a Interschool Science festival held at Chinmaya Vidyalaya,Virugambakkam,Chennai. I was Invited to Judge the Robotics event where young innovators from various school from chennai participated.
This paper describes about the Martian Exploratory Rover which can reduce the various barriers in surface exploration. The Rover is inspired from the anatomy of an ant. Like an ant’s body the rover consist of individually suspended six leg mechanism which enables the rover to move through rough terrains. The Dinoponera agility is implemented in a carbon fibre Suspension. The chassis are controlled by automatic feedback system that allows the rover to maintain traction and with the alternating thread pattern in the wheels, it can travel through different kinds surfaces. The rover carries a series of small Swarm bots which are loaded on the rear end of the main rover. The Swarm bots are designed to explore narrow caves and dents to gather soil samples. The processing is faster as the Swarm bots are attached to the Main Rover’s Network. vThe analysing technique is carried out with A* algorithm coupled with Artificial Intelligence and Machine Learning in order to optimize Efficiency and Fuel Consumption.
Metaplore enables enterprises develop awareness on the threat perceptions of technology risks. We prepare a strategic roadmap through earmarking the areas of technical risks that include infrastructure, application, systems and process risks.
Armed with the best skilled resources, we make sure that internal controls are put in place across the IT environment. Our IT Auditing experts have a sound knowledge in internal information system audits, security services and application control, besides pre- and post-implementation reviews.
I worked with opencv and c++ for security services.
Martian Exploratory Rover designed with the suspension system based on the anatomy of a spider. The Arachnid’s agility is implemented in the suspension structure. This supports the chassis in an automated feedback hydraulic system that allows the Rover to maintain optimum Centre of Gravity position regardless of the obstacle it traverses Utilizing the Rover’s belly cameras the Wheels are protected from contact with foreign objects that pose a potential threat to the movement of the Rover. The Guidance system of the Rover is automated, and relies upon stereo imaging from the lead camera. This analyses 3 ft. in front of the rover, and carries out an Algorithm based upon the A* path finding algorithm coupled with Fuzzy Logic to optimize efficiency of traversal and fuel consumption. The Robotic Arm onboard the Rover is a 7 DOF Robotic Arm. The robotic arm is also equipped with a rotating auger in order to collect soil samples for on-site analysis aboard the chassis. Sensors and instruments to identify the presence of sugars, amino acids, lipids are integrated on the rover.
In this paper, we have designed a autonomous vehicle which is cost effective and powered by Robotic Operating System (ROS). The vehicle is capable of maintaining a constant speed and distance for monitoring or surveillance. ROS is implemented for trajectory tracking and telemetry. A low cost compact on-board embedded system powers the vehicle. Various image processing techniques are been implemented for navigation and obstacle detection. Artificial Neural Network which helps in finding the shortest path by using the acquired data from image processing. Different controllers were implemented for movement and obstacle avoidance including PI and PID. The performance were compared and the results are also discussed in this paper.
In this paper, we have designed a quad copter which is cost effective and powered by ROS. The drone is capable of maintaining a constant altitude during hovering and a constant position for monitoring or surveillance. Robotic Operating System is implemented for trajectory tracking and telemetry. A low cost compact on-board embedded system powers the drone. Different controllers were implemented for position and altitude lock including PI and PID. The performance were compared and the results are also discussed in this paper. the results are also discussed in this paper.
Hindustan Mars Rover Team aims to design and build the next generation of Rover and its associate Robots which aids Humans for Space Exploration.As The Technical Manager of the team my role was to monitor and guide Electronics , Software and mechanical team.
FLYTTA is a NEXT GENERATION MOBILITY Platform, offers Hassle free logistics and exceptional level of transparency to the customer using Existing Supply Chain, Deep Learning and Vision Engineering.As a Technical Manager for the company i was responsible for the management of the de... See more
Fully Autonomous Low Cost Drone works on fighting fire where man cannot set foot. There are some cases where too many firefighterssacrificed their lives to help one soul live. Fire in narrow crooked places may result in the accumulation of debris inside which fire may sustain and average sized fire-fighters cannot enter.
Ucal-Jap is a Uav Design and manufacturing company in India.
Development of an autopilot system for UaV using image processing with ML.
Worked on technologies such as Machine learning , image processing , Arduino.
Hindustan Mars Rover Team aims to design and build the next generation of Rover and its associate Robots which aids Humans for Space Exploration. Being in the first year into the team and second year of my college i was a software developer for the team. We used Robot operating System with c++ . We used arduino microcontroller for control system and Qt framework For UI development