After 2012 when deep learning based techniques won the ImageNet contest with a clear margin to competing algorithms, deep learning has been called “the revolutionary technique that quietly changed machine vision forever”. For many classification tasks deep learning has drastically surpassed previous state of the art results in classification accuracy. Currently large deep neural networks achieve the best results on speech recognition, visual object recognition, character recognition, and several language related tasks.

The power of deep learning

Deeper machine learning architectures are better capable of handling complex recognition tasks compared to previous more shallow models. Major benefits of deep networks are:

• their superior modeling capabilities of heterogeneous data in layers of increasing complexity,
• their ability to learn the best features to represent the raw data, and
• their ability to gain in performance with the availability of more training data.
   

Specialized deep learning

The next wave of vision technology will take place for other applications than traditional computer vision, such as medical imaging, marine and seismic imaging, remote sensing of ocean, land and infrastructure, process monitoring and industry. These applications depend upon non-standard imagery and present challenges that needs to be solved in order to benefit from the untapped potential:

• Learning from limited data sets
• Transferring knowledge across domains
• Exploiting non-standard and heterogeneous imagery
• Capturing context and dependencies
• Quantification of uncertainties in predictions
• Reliable and explainable predictions
   

Combining years of experience in image analysis and machine learning

The image analysis and machine learning group at NR and the machine learning group at UiT work together to better understand the needs and to develop state-of-the-art specialized deep learning solutions suitable for solving specific problems for various industry-, medical and environmental applications.

Current activities and projects involving deep learning
 

	MIM Norwegian Cancer Registry, partly funded by the Research Council of Norway Use of deep learning and Big Data in the Norwegian Breast Cancer Screening Program
	DELI Equinor Deep learning for seismic interpretation
	COGSAT European Space Agency Automatic analysis of Sentinel-data using deep learning techniques
	COGMAR IKTpluss, Research Council of Norway Automatic analysis of marine data using deep learning techniques.
	InfraUAS Orbiton AS, partly funded by the Research Council of Norway, BIA programme Monitoring of critical infrastructure using UAVs
	INCUS GE Vingmed Ultrasound, partly funded by the Research Council of Norway, BIA programme Intelligent Cardiovascular Ultrasound Scanner
	Hyperbio TerraTec, partly funded by the Research Council of Norway, BIA programme Automatic mapping of forest species using deep learning
	AIRQUIP Research Council of Norway Automatic estimation of traffic from VHR satellite images using deep learning techniques.
	HBR Infrastructure programme, Research Council of Norway Transcription of historical Norwegian census forms
	UAVSEAL Institute of Marine Research, funded by the Research Council of Norway Detection and counting of seals on ice from aerial images
	LASTRAK Norwegian Mapping Authority Tracking of small roads and forest paths from laser data
	CULTSEARCHER Norwegian Directorate for Cultural Heritage Detection of cultural heritage sites from laser data
	SNOWBALL EEA grants Automatic detection and mapping of avalanches in optical and SAR satellite images.