CTR Carinthian Tech Research AG is an industry-oriented research and development center for „Smart Sensors and Systems Integration“. As the largest non-university research center in southern Austria, CTR has gained a reputation for expertise in R&D sensor technologies serving science and industry at both national and international level. CTR focuses on four main research areas: Microsystem Technologies, Heterogeneous Integration Technologies, Photonic Sensor Systems and Smart Systems. CTR is part of Austria’s COMET research programme with the K1 excellence center “ASSIC Austrian Smart Systems Integration Research Center”..
Deep learning approaches have become the state of the art for object classification and detection [He et al., 2015; Ren et al., 2016]. Recently, deep learning has been applied successfully for depth map predictions from monocular images [Eigen et al., 2014; Kuznietsov et al., 2017].
This diploma thesis will evaluate different deep learning/machine learning approaches for advanced image analysis such as volume estimation of objects from monocular images. In a first step, the best methods for recording the ground-truth data will be investigated and implemented. In a next step, different network models and architectures will be trained and evaluated with the data obtained. Data preprocessing and custom network layers will be implemented in Python (or C++/Cuda, if needed).
The prospective candidate is encouraged to present the results at a national or international conference and/or publish the results in a scientific journal.
- Create and prepare dataset for training and testing.
- Develop NN architectures optimized for volume estimation.
- Test and evaluate the different architectures with respect to the task at hand.
- Writing and defense of the thesis.
- Publication of the results at a conference or in a scientific journal.
Start Date / Duration / Contract
- Start date (planned): immediately
- Duration (planned): 6 months
- Contract: Payment is based on the collective contract.
- Place: Villach
- Student in one of the following fields: statistics, machine vision, machine learning, mathematics, informatics
- Experience: object recognition, learning algorithms, software implementation, Python or Matlab, C/C++/Cuda (desired but not required)
- Experience with popular deep learning frameworks desired (Caffe, TensorFlow, etc.)
- English: fluent, German: at least basic knowledge desirable
Contact: Jan Steinbrener Email: email@example.com
Please fill our application form www.ctr.at/en/application.