Call for Applicants for the TUMO Labs + AOByte Artificial Intelligence Project
Mandatory: basic knowledge of Python, OOP and data structures, linear algebra,
Optional: basic knowledge of neural networks and optical recognition
Duration: 3 months, 3 days weekly, 3 hours within 9 am and 3 pm
Application Deadline: January 20, 2020
Start Date: February 3, 2020
Cost: free of charge
Eligible: students or non-students with the required knowledge, 18+
AUA and UFAR students will receive a university credit
Objective: Building an Easy Operation Tool for Traffic Control
The project consists of two modules, an AI-powered car license plate recognition module and a back-office web app that allows the user to moderate which cars can pass through a checkpoint. The system contains the following: a number-plate recognition system and back office API. Security access is performed through the number-plate recognition system. The basic workflow of the system is: following a vehicle approaching a security checkpoint, performing optical recognition to detect the vehicle’s plate number, and checking the detected plate number against the checkpoint’s database to determine whether the vehicle can pass the checkpoint on this date and time.
What Will Students Learn?
In the scope of this project, students will work on an AI module. AOByte will already have the fundamental structures of the project implemented (an AI-powered car license plate recognition module) before students begin. Students will be engaged in collecting and preprocessing the data, training the model based on the gathered data and, finally, evaluating the model. Students will also assist AOByte in optical recognition cases.
Students will gain knowledge and experience in working in a team while gaining practical skills in machine learning and software development.
The project consists of 3 primary parts:
Introduction, project setup, theoretical basics of machine learning, data gathering/preprocessing.
Students will gain on-hand experience in constructing artificial neural network (ANN) with TensorFlow and Keras libraries and will be able to choose appropriate ANN architecture for the project.
Training, testing and evaluating.
Over the following six weeks, the group will implement CNN models for number-plate recognition. During this step, students will work on improving existing models provided by the AOByte team. Students will gain knowledge in how to train and test the model and how to debug a machine learning algorithm.
Improving the accuracy of the model and finalization/deployment of the project.
During the final step, students will try to improve the accuracy of the model and test it in real life. Students will deepen their skills in the fine-tuning of CNN and will gain knowledge in model deployment․