Pierrick Tranouez
Document Actions
|
|
To the left a collection of Boids-like individuals; to the right the dynamic perception of their flocks.
More details in [Tra 09]
Ancient newspapers and medieval codices
|
|
03/10/2012
I have been working in two research projects aiming at helping historical documents enter the digital age. The first is included in the PlaIR project, and is focuses on improving accessibility to the archives of Le Journal de Rouen, one of France eldest local newspaper: it was published from 1762 to 1947. We have tried to push beyond the too usual "let's just put our scans online, it will be good enough", by providing more confortable functionalities:
- Automatic transcription (OCR)
- Articles detection.
- Full text indexing
- Collaborative corrections of the transcription mistakes (crowdsourcing)
- Annotating and tagging the articles.
We have a demo here, and more detailed explanations in our DocEng paper.
Our second project is DocExplore. The aim is similar, but with older and more diverse documents. We created a software suite into which you may input your collection of digitized manuscripts. The software provides basic collection management tools, but its strength is in the possibility to create Region Of Interests. ROI are zones of an image, of any shape and number, to which are associated augmenting data: text (transcription, annotations), image, video, hyperlink (including to other ROI).
From this collection of augmented pages, non-linearly connected, a virtual book can be extracted, mapped on a 3D representation, which can be browsed classically or through a tactile or gestural interface.
JNSA
|
|
28/09/2006
We wrote a new article on urban traffic simulation, with an additional angle of simulating emergency or crisis situations, in the Journal of Nonlinear Systems and Applications (JNSA). It's so far probably the best article on what we did, less detailed but more synthetic than the ones in ICCSA. We describe how we build a graph then a trafic network from shapefiles, so as to have a geographically correct simulation. We explain the way we modeled each car with an agent, with a quite sophisticated behavior, reactive when everything is fine, switching to more cognitive when it's not. Despite the details, it runs faster than real time on a real urban agglomeration, with tens of thousand of vehicles. We finally describe the new family of tactics aimed at modeling the behavior of drivers in times of crises.
An advanced draft can be read here.
