Computer vision and AI
Stavros Tsogkas
So what is Computer Vision?
I still don't get it...
People don't just "see"
other cars
how far?
Don't run them over!!
which traffic light?
Goal: make sense of visual data
Semantic segmentation (Cityscapes dataset)
road
car
person
sidewalk
traffic light
vegetation
pole
building
Fully convolutional neural networks
P(person)
P(horse)
:
P(dog)
dog
person
- Fully convolutional networks for semantic segmentation, Long et al., CVPR 2015.
- Semantic image segmentation with deep convolutional networks and fully-connected CRFs, Chen et al., ICLR 2015..
Part segmentation in images
Deep learning for semantic part segmentation with high-level guidance, Tsogkas et al., arxiv
Medical applications
Alzheimer's:
structure degeneration
Schizophrenia: volume abnormalities
[Shenton M.E. et al., Psychiatry Res. 2002]
Tumors: avoid radiation on sensitive regions
[Hoehn D. et al., Journal of Medical Cases, 2012]
Why automatic segmentation?
Putamen
Ventricle
Caudate
Amygdala
Hippocampus
Visualization and inspection
No need for manual annotation
(time consuming, need experts,
limited reproducibility)
Non-invasive diagnosis and treatment
Part segmentation in FMRI
Subcortical brain structure segmentation using FCNNs, Tsogkas et al., ISBI 2016
From 2D slices to 3D volumes
Our results
Groundtruth
More computer vision at UofT
- Object recognition (Sven Dickinson)
- Vision and language (Sanja Fidler)
- Computational photography (Kyros Kutulakos)
- Electron Cryomicroscopy (David Fleet)
- Semantic segmentation (Raquel Urtasun)