Pol Moreno Comellas and Aleksej Stolicyn
Tue 04 Apr 2017, 11:00 - 12:00
IF 4.31/4.33

If you have a question about this talk, please contact: Gareth Beedham (gbeedham)

Pol Moreno Comellas

"Overcoming Occlusion with Inverse Graphics"

Scene understanding tasks such as the prediction of object pose, shape, appearance and illumination are hampered by the occlusions often found in images. We propose a vision-as-inverse-graphics approach to handle these occlusions by making use of a graphics renderer in combination with a robust generative model (GM). Since searching over scene factors to obtain the best match for an image is very inefficient, we make use of a recognition model (RM) trained on synthetic data to initialize the search. This paper addresses two issues: (i) We study how the inferences are affected by the degree of occlusion of the foreground object, and show that a robust GM which includes an outlier model to account for occlusions works significantly better than a non-robust model. (ii) We characterize the performance of the RM and the gains that can be made by refining the search using the GM, using a new dataset that includes background clutter and occlusions. We find that pose and shape are predicted very well by the RM, but appearance and especially illumination less so.  However, accuracy on these latter two factors can be clearly improved with the generative model.




Aleksej Stolicyn

"Computational Etiology of Cognitive Deficits in Depression"

Depression is a prevalent psychiatric condition with a high level of economic impact. A large fraction of this impact is due to lost productivity. Consistent with high productivity costs, depression is also associated with a range of cognitive deficits. Despite their prevalence, etiology of these deficits still remains not clear. In this talk I will first outline a descriptive framework which links motivational factors with cognitive performance, based on several recent theories and reviews of neural bases of motivation and cognition. I will then present computational (algorithmic) models of two cognitive tasks, consistent with this general framework. Finally, I will show how introduction of etiological factors of depression – low estimated task control, altered positive and negative valuation – can replicate deficient performance at the tasks. The models serve as a computational and algorithmic bridge between etiological factors of depression and cognitive deficits, and show that different depressive factors can lead to different patterns of behavioural performance.