Optisol Business Solutions proudly presents “Madurai Thamizhukku oru AI” online Hackathon for students and professionals across the globe.
Scope of the event is to build a mobile application leveraging Artificial Intelligence and Machine learning to perform OCR on Tamil language

Tamil, one of the longest surviving classical languages in the world, spoken natively by the people of south Asia, and considered as official language in Tamil Nadu, Srilanka, Singapore and Pondicherry. The earliest period of Tamil literature, Sangam literature, is dated from 300 BC – AD 300. It has the oldest extant literature among Dravidian languages. With artificial intelligence and computer vision aiding humans ease out almost every task. This hackathon focuses on implementing optical character recognition on Tamil language.

Tourists who come visit Tamil Nadu have hard time exploring the beauty of Tamil Culture if they don’t know Tamil. This is because mostly (not always) everything ranging from road signs, route boards on buses and trains to menus in hotels are written in Tamil. They would also have a hard time communicating to village localities as people in our native lands could merely understand foreign accent/language (but this is out of our scope). Our scope in this hackathon is to mitigate this difficulty in reading signs in Tamil. We expect a solution which involves computer vision, android and cloud tech stacks. The following is a detailed description of our expectations.

• Computer Vision: Optical character Recognition of Tamil language from the images of road sign board, bus/train route boards, hotel menus, books etc. The trained model/logic should be packed as an API. The API should be capable of reading a text from GET request, perform OCR to extract the characters and translate them to English language. The result data should be passed as a text by the API.
• Cloud: The OCR solution with API should be hosted on cloud and the API should be accessible when hit from remote.
• Android: A mobile application that could run on android should be built. This application should be able to capture or select an image from local storage to run OCR on. The Image should be passed to the Cloud API which will return the translated text after processing. The text should be displayed in the mobile application.

Guidelines:

• Any data/images that would be required should be collected by the candidates. Optisol will not provide any image set for training and development.
• Candidates are free to use any cloud service provider of choice and are advised to use functionalities that are provided by its free tier.
• The performance of the solution will be tested based on accuracy, throughput and scalability of the solution.
• Candidates are free to use any open-source packages/modules.
• Past the mid-way of the hackathon, there would be a check-in where the panel members would assess the progress of the teams. Candidates will be provided with a sample of the test set on which the inference will be tested on.
• The candidates will be asked to submit the source code, API that was hosted on cloud and .apk file of the mobile application. Proper coding standards would provide an edge during evaluation.
• On the final day of hackathon, candidates will be given a test set of images which they have to provide as an input to the mobile application and record a demo with it. This demo should be submitted along with other artifacts.