Browse Poster Wiki: 2022 Poster Award Winners, Awards Ceremony (Watch Recording from minute marker 1630), Posters by 2022 HPCC Systems Interns, Posters by Academic Partners, Poster Judges, About Virtual Judging, 2022 Poster Contest Home Page, Poster Contest Previous Years

Shrikar Swaroop is a 2nd Year Student at the RV College of Engineering, Bangalore, India.

Poster Abstract

The past decade has seen a rapid increase in multimedia generation. Consequently, novel algorithms have taken shape during this time which try to make sense of these multimedia and provide a means to analyze data on a large scale. Single Shot Detection is one such approach. The Single Shot Multibox Detector (SSD) is an object detection algorithm that leverages Deep Learning to detect various classes of objects in input images. It localizes and classifies probable objects by using bounding box regression. The detection is in the form of closed boxes around probable objects, hence the name Multibox. The model itself can be briefly explained as follows:

o Single Shot: Localization and classification only takes a Single Forward Pass.

o Multibox: The bounding box regression technique used by the model.

o Detector: The model is an object classifier and detector.

HPCC and SSD

SSD deals with detecting objects within images, thus the inputs i.e., images can be entirely different and unique each time. For such a model to be successful and efficient, it must be trained on a large imagery data set that is both robust and represents a diverse range of real-life scenarios. Further, the model’s characteristics are solely dependent on the training data set being used. Thus, such models overlap into the big data territory and requires an efficient means to filter, clean, and manage data. Here, the services and features offered by HPCC can be leveraged to create more robust models that can span many real-life scenarios.

Why SSD

SSD was one of the pioneering models that were developed to tackle object detection. By mapping this model using HPCC, a basic framework/baseline can be developed which can further be used to explore more complicated models for big data applications using HPCC

Presentation

In this Video Recording, Shrikar provides a tour and explanation of his poster content.

Single Shot Multibox Detection

Click on the poster for a larger image.