BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.2.3.1//EN TZID:Asia/Kolkata X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT UID:40@cds.iisc.ac.in DTSTART;TZID=Asia/Kolkata:20240301T150000 DTEND;TZID=Asia/Kolkata:20240301T160000 DTSTAMP:20240221T085040Z URL:https://cds.iisc.ac.in/events/m-tech-research-thesis-colloquium-cds-sc alable-read-alignment-algorithm-for-cyclic-pangenome-graphsscalable-read-a lignment-algorithm-for-cyclic-pangenome-graphs/ SUMMARY:M.Tech Research Thesis {Colloquium}: CDS : "Scalable Video Data Man agement and Visual Querying for Autonomous Camera Networks" DESCRIPTION:DEPARTMENT OF COMPUTATIONAL AND DATA SCIENCES\nM.Tech Research Thesis Colloquium\n\n\n\nSpeaker : Ms. Bharati Khanijo\n\nS.R. Number : 06 -18-02-10-12-19-1-17219\n\nTitle :"Scalable Video Data Management and Visu al Querying for Autonomous Camera Networks"\nResearch Supervisor: Prof. Yo gesh Simmhan\nDate &\; Time : March 01\, 2024 (Friday) at 03:00 PM\nVen ue : # 102 CDS Seminar Hall\n\n\n\nABSTRACT\n\nVideo data has been histori cally known for its unstructured nature and rich semantic content but also for scalability issues in terms of storage and analytics. Mobile aerial p latforms like drones capture such videos across space and time. Advances i n computer vision and deep learning enable automatic extraction of rich se mantic information from video data\, leading to applications where the sto red video data can be used to study and analyze the world retrospectively and automatically. However\, recent research has highlighted the compute-i ntensive nature of such Deep Neural Network (DNN) models\, e.g.\, for accu rate object detection\, leading to high computing costs that limits their applicability for brute-force analysis of all historical videos. Also\, an efficient design of such applications often requires co-analysis of video data along with associated geospatial and temporal metadata\, which is a challenge.\n\nWe propose a geospatial-temporal video query system with sup port for semantic queries for drone videos\, extending an existing spatial -temporal database and contemporary object detection models. We develop a heuristic to enable better reuse of semantic object detections obtained fr om different configurations (object detection model and its input resoluti on) . The system further motivates the need for optimizations for retrospe ctive semantic analysis and storage for drone videos\, which is addressed by our novel DDownscale method and the associated ingest pipeline.\n\nPrio r optimizations on semantic querying over video data focus on static camer as from city-scale traffic/surveillance camera networks\, often leveraging the spatial and temporal characteristics of associated videos\, which are absent in videos recorded by mobile drone cameras. We specifically focus on two such characteristics of drone videos. One is that drone videos have shorter durations\, unlike those captured by static cameras. Another is t hat there can be large variations in the level of detail of information ca ptured across a fleet of drone cameras due to differences in the resolutio n of the camera\, the altitude\, and the orientation from which the videos were captured.\n\nSpecifically\, we address the need to intelligently sca le-down the spatial resolution of videos to reduce the video storage costs and semantic query/inferencing time. However\, conventional methods of ma nual or profiling-based estimation of the ideal scaling ratio are compute- intensive and/or time consuming for such heterogeneous feeds. We propose D Downscale\, a novel method to dynamically select the downscale factor for a video by utilizing the information on the object size in the video. We m odel the downscale factor and associated drop in relative recall due to do wnscaling as a function of object size in the downscaled video and demonst rated that for a given DNN model and class of interest\, DDownscale genera lizes well to the evaluated datasets. A DDownscale inequality between the relative recall drop and the hyper-parameters of the method is derived. Th is satisfies 98% of the dynamically downscaled videos across datasets\, ob jects of interest and parameters. The algorithm achieve over 19% reduction in total object detection time and 24% reduction in storage on average co mpared to the baseline of storing/inferencing at the original resolution \ , for different user-specified target reduction in recall values ranging f rom 1--30%\, and 96% of the downscaled videos are within the target recall drop.\n\nA simpler specification at the time of ingest of target level of detail (average ground spatial distance) captured in the video and the ha rmonic mean of relative recall drop for the class of smallest object of in terest and selected object detection model was derived using the above mod eling to aid in the selection of a target level of detail. Additionally\, we develop an ingest pipeline that reduces the time to ingest drone videos using this dynamically downscaling method over heterogeneous edge acceler ators\, and reduce the average turnaround time to ingest data from multipl e clients by ~ 66%\, despite the downscaling time overhead\, compared to u ploading original resolution video without downscaling.\n\n\n\nALL ARE WEL COME CATEGORIES:Events,MTech Research Thesis Colloquium END:VEVENT BEGIN:VTIMEZONE TZID:Asia/Kolkata X-LIC-LOCATION:Asia/Kolkata BEGIN:STANDARD DTSTART:20230302T150000 TZOFFSETFROM:+0530 TZOFFSETTO:+0530 TZNAME:IST END:STANDARD END:VTIMEZONE END:VCALENDAR