Master Thesis "Automatic annotation tool for accurate object recognition using Convolutional Neural Networks"
High Tech Campus Villach, Europastraße 12, 9524 Villach, Austria
Your future responsibilities
- Accurate object recognition in a sequence of images is a challenging problem. The state-of-the-art algorithms for object recognition rely on Convolutional Neural Networks, which however require a lot of data for the training phase. There is then the need for developing tools to augment and speed up the annotation process, in order to build larger labelled datasets.
- Open source annotation tools exist, but they always require manual intervention from a human user. The possibility of implementing semi-automatic annotation methods has the potential for greatly reducing the load of such process. At SAL our goal is to perform automatic annotations with little human intervention building a system-in-the-loop using Convolutional Neural Networks.
- The goal of the master thesis is to implement the semi-automatic annotation tool, train it on acquired data and test its performance.
In particular, the candidate will:
- Design a tool for semi-automatic object annotation in image sequences, making use of Convolutional Neural Networks as object detector.
- Acquire data to be annotated.
- Develop the tool and train the object detector on the acquired data.
- Write and defend the thesis, optionally write a scientific publication.
- Master student in Engineering, Computer Science or related fields.
- Experience with Machine Learning and Computer Vision is also beneficial.
- Autonomous, well organized and enjoying work in a dynamic and multicultural team.
- Oral and written communication skills in English.
- Start date (planned): Spring/Summer 2020
- Duration (planned): 6 Months
- Payment: 1.500 € monthly gross (38,5 hours per week)
- Place: Silicon Austria Labs, Villach, Austria
- State-of-the-art laboratories and equipment.
- An international scientific environment and contacts with our industrial network.
- Opportunity to work on a hot topic with industrial collaborations.