ILA GOKARN

I am a fifth-year PhD candidate in Computer Science, advised by Prof. Archan Misra, at the School of Information Systems in Singapore Management University. Currently, I work primarily in the field of pervasive systems and sensing with a focus on cognitive edge computing paradigms and platforms.

I received my Bachelor in Science in Information Systems at Singapore Management University in 2015. Prior to starting the PhD program, I spent 4 years gaining industry experience as a Software Systems Engineer at Arista Networks and Cisco Systems. I specialised in software systems, cloud management platforms, edge computing, and IoT ecosystems for smart cities. 

[Resume] (October 2022) [LinkedIn]   [Google Scholar] [Research Statement] (August 2019)

RECENT HIGHLIGHTS

[April 2024] I'm thrilled to share that I  have been selected as a MobiSys 2024 Rising Star 🌟

[April 2024] Our poster "Profiling Event Vision Processing on Edge Devices" has been accepted for publication at MobiSys 2024.

[March 2024] Our paper "JIGSAW: Edge-based Streaming Perception over Spatially Overlapped Multi-Camera Deployments" has been selected for publication at ICME 2024. 

[January 2024] Our paper  "Algorithms for Canvas-based Attention Scheduling with Resizing" has been accepted for publication at RTAS 2024.

[January 2024]  Our demo titled "Demonstrating Canvas-based Processing of Multiple Camera Streams at the Edge" won the "Best Demo Award" at COMSNETS 24!

[December 2023] Our demo paper titled "Demonstrating Canvas-based Processing of Multiple Camera Streams at the Edge" has been accepted for publication at COMSNETS 24.

[August 2023]  Our chapter, "Lightweight Collaborative Perception at the Edge," is now published by Springer in the book "Artificial Intelligence at the Edge".

[June 2023] I interned with the Pervasive Systems Research Group at Nokia Bell Labs for the summer of 2023. 

[May 2023] Our work titled "Underprovisioned GPUs: On Sufficient Capacity for Real-Time Mission-Critical Perception" has been accepted for publication at ICCCN 2023.

[March 2023] Our work titled "MOSAIC: Spatially-Multiplexed Edge AI Optimization over Multiple Concurrent Video Sensing Streams" has been accepted for publication at ACM Multimedia Systems (MMSys) 2023.

[July 2021] We are presenting our work "VibranSee: Enabling Simultaneous Visible Light Communication and Sensing" at SECON 2021. 

[June 2021] Awarded the N2Women Young Researcher Fellowship for SECON 2021. 

[March 2021] We are presenting a short paper on Adaptive & Simultaneous Visible Light Sensing and Communication at PerCom 2021. 

[January 2021] Awarded "Best Research Demo Award" at COMSNETS 2021 for our work "Demonstrating Simultaneous Visible Light Sensing and Communication"

Research

JIGSAW's Overall Functionality and Target Application: (a) Multiple spatially overlapped cameras deployed at a traffic intersection are processed by a single edge device (b) At DNN inference time T_i, estimated regions of interest (tiles) are extracted at their appropriate scale from the source camera frames and (c) evaluated for their utility to the streaming perception task before being packed onto a canvas frame.

JIGSAW: Edge-based Streaming Perception over Spatially Overlapped Multi-Camera Deployments

We present JIGSAW, a novel system that performs edge-based streaming perception over multiple video streams, while additionally factoring in the redundancy offered by the spatial overlap often exhibited in urban, multi-camera deployments. To assure high streaming throughput, JIGSAW extracts and spatially multiplexes multiple regions of interest from different camera frames into a smaller canvas frame. Moreover, to ensure that perception stays abreast of evolving object kinematics, JIGSAW includes a utility-based weighted scheduler to preferentially prioritize and even skip object-specific tiles extracted from an incoming stream of camera frames. Using the CityflowV2 traffic surveillance dataset, we show that JIGSAW can simultaneously process 25 cameras on a single Jetson TX2 with a 66.6% increase in accuracy and a simultaneous 18x (1800%) gain in cumulative throughput (475 FPS), far outperforming competitive baselines.

Illustration of MOSAIC’s Overall Functionality: (a) Input frames captured by cameras (b) Packing tiles from multiple images onto a single canvasframe (image not to scale).

MOSAIC: Spatially-Multiplexed Edge AI Optimization over Multiple Concurrent Video Sensing Streams

Sustaining high fidelity and high throughput of perception tasks over vision sensor streams on edge devices remains a formidable challenge, especially given the continuing increase in image sizes (e.g., generated by 4K cameras) and complexity of DNN models. One promising approach involves criticality-aware processing, where the computation is directed selectively to ``critical" portions of individual image frames. We introduce MOSAIC, a novel system for such criticality-aware concurrent processing of multiple vision sensing streams that provides a multiplicative increase in the achievable throughput with negligible loss in perception fidelity. MOSAIC determines critical regions from images received from multiple vision sensors and spatially bin-packs these regions using a novel multi-scale Mosaic Across Scales (MoS) tiling strategy into a single `canvas frame’, sized such that the edge device can retain sufficiently high processing throughput.  Experimental studies using benchmark datasets for two tasks,  Automatic License Plate Recognition and Drone-based Pedestrian Detection, shows that MOSAIC, executing on a Jetson TX2 edge device, can provide dramatic gains in the  throughput vs. fidelity tradeoff. For instance, for drone-based pedestrian detection, for a batch size of 4, MOSAIC can pack input frames from 6 cameras to achieve (a) 4.75x (475%) higher throughput (23 FPS per camera, cumulatively 138FPS) with <=1% accuracy loss, compared to a First Come First Serve (FCFS) processing paradigm.

[PDF] 

Example use case of machine condition monitoring where a factory floor robot is performing sensing of machine vibrations using VLS while simultaneously sending control commands to the machine using VLC, thus closing the IoT control-feedback loop using visible light.

VibranSee: Adaptive, Simultaneous, Visible Light Communication and Sensing

Visible Light Communication (VLC) goodput (or application-level throughput), and Visible Light Sensing (VLS) accuracy or coverage demonstrate a natural trade-off depending on the duty cycle of the light source. Intuitively, VLS ideally assumes the use of a strobing source with an infinitesimally small duty cycle, whereas VLC goodput is directly proportional to the active period of each individual pulse, maximized when the duty cycle is 100%. We used this understanding to design mechanisms that moderate this tradeoff – a time-multiplexed single strobe architecture and a harmonic multi-strobe architecture. Based on these understandings, we designed VibranSee, an adaptive mechanism that further fine-tunes the tradeoff between VLC and VLS, and setup experiments on cheap commodity pervasive devices - Arduino and Raspberry Pi. 

[PDF] [Demo]

Selected conference Publications

Other publications

Book Chapters

Demonstrations and Posters

Honours and Awards

People

Affiliations

Nokia Bell Labs

University of Illinois Urbana- Champagne

Living Analytics Research Center

Singapore Management University

Arista Networks

Cisco Systems

Personal

I am a trained Bharatnatyam dancer and I am now pursuing the Odissi form as well with Ethos Odissi. I am actively involved in the fine arts community in Singapore. I am also involved with mentoring young girls interested in STEM research and industry.

contact

Reach me at

ingokarn(dot)2019(at)phdcs(dot)smu(dot)edu(dot)sg


Living Analytics Research Center 

School of Information Systems

Singapore Management University

80 Stamford Rd