- Project name
- Deep learning method for hyperspectral image classification
- Mohammadmehdi Saberioon, PhD.
- Description of the activities of the project
- Hyperspectral imagery system as emerging technology that integrate both spectroscopy and imaging in a single system has potential to capture the subtle spectral difference under different physiological and pathological condition. The output of the hyperspectral imaging system is a three-dimension (3D) form (spatial-spectral-spatial) and is referred to as spectral cube, data cube, spectral volume or simply hybercube. Hybercube is consist of a large amount of information could be used to analysis and ascertain minor and/or subtle physical and chemical features in samples
- Deep learning based methods achieving promising performance in many fields. Deep learning has recently outperformed some other machine learning methods on many vision related tasks including image classification, object detection, scene labelling and face recognition.
The main aim of this project is evaluating the feasibility of deep learning methods for classifying hyperspectral images. The student will develop a deep learning algorithm based on convolutional neural network to implement the classification of hyperspectral images for discriminating fish fillets based on the different diets during rearing cycle.
- Number of students - 2
- secondary school students - NO
- university students - YES
- Prepositions: Python or The R programming, basic image processing
Academic and University Center
Zamek 136, Nove Hrady