Jessica Edwards is a sophomore at Harvard College concentrating in Computer Science and intending to get a secondary in Educational Studies. This past summer she was a Research Assistant at Johns Hopkins University's Human Language Technology Center of Excellence and is currently participating in computer science research term-time through the Learning, Innovation, and Technology Lab in Harvard's Graduate School of Education. At Harvard, she is also involved in Harvard's Women in Computer Science organization, volunteers at Y2Y Harvard Square Homeless Shelter, and is a member of the Kuumba Singers of Harvard College.
Multilingual Text Search in Image Collection
This talk will focus on computer vision techniques used to explore core aspects in the optical character recognition (OCR) pipeline, specifically focusing on text localization. I will explain how the team I worked with helped make the Efficient and Accurate Scene Text detector (EAST) more accurate by running models on more data, focusing on how synthetic data, real data augmentation, and generative adversarial networks (GANs) helped in this process. I will also talk about my experience doing research this summer in the Summer Camp for Applied Language Exploration (SCALE) program as well as attending workshops through JHU’s Center for Language and Speech Processing (CLSP).