Direkt zum Inhalt springen
login.png Login join.png Register    |
de | en
MyTUM-Portal
Technical University of Munich

Technical University of Munich

Sitemap > Bulletin Board > Diplomarbeiten, Bachelor- und Masterarbeiten > Master or Bachelor Thesis / IDP / FP: Machine Learning for Human Intention Recognition from RGBD video
up   Back to  News Board    previous   Browse in News  next    

Master or Bachelor Thesis / IDP / FP: Machine Learning for Human Intention Recognition from RGBD video

15.05.2020, Diplomarbeiten, Bachelor- und Masterarbeiten

Overarching goal of this thesis is to take state-of-the-art Machine Learning algorithms for estimating Human Intentions to the next level.

This thesis will be carried out at the EI Chair of Media Technology (Prof. Steinbach)

Human Activities of Daily Living are driven by our underlying intentions. For example, the Intention of "making Pasta" spawns a sequence of activities like fetch pasta, boil it, fetch and chop vegetables for the sauce, and clean up after cooking.

Correct estimation of human intentions is critical for non-intrusive and anticipatory assistance in these tasks by robots. Based on sensory observation of human activity, Machine Learning models have been built for estimating human intentions. State-of-the-art models are based on Conditional Random Fields (CRFs), which are a well-known tool for capturing patterns in sequential data. But CRFs have an important drawback. To keep inference computationally tractable, usually only first order dependencies between adjacent data segments/labels are considered.

Daily human activities, on the other hand, usually feature long-distance "higher order" dependencies, e.g. an object fetched from a cupboard will probably be returned back to its place sometime after use. This thesis should investigate approaches to model higher order dependencies for improving inference, while exploiting sparsity in real-world data to keep the problem computationally tractable.

We will begin with publicly available datasets, e.g. Cornell CAD-120: http://pr.cs.cornell.edu/web3/CAD-120/. Self-recorded data may be used if required at a later stage.

Kontakt: rahul.chaudhari@tum.de

More Information

https://www.ei.tum.de/lmt/team/mitarbeiter/chaudhari-rahul/studentische-arbeiten/

Todays events

no events today.

Calendar of events