The main goal of the Cognitive Vision project is to improve the results of artificial intelligence. The current scope is to equip the abilities of the AIBO (robot dog of the Sony) to explore its environment and to recognize the local position (e.g. navigation in a room). The first trial is to use and develop image processing methods with a single webcam and to apply these techniques with AIBO.
PED is a dialogue management system that uses a probabilistic nested belief model to choose dialogue strategies. The dialogue system designer need only supply a set of plan rules to PED as a dialogue grammar with preconditions. Using this grammar, PED constructs game trees (like the one below) to represent the outcomes of the dialogue, so that a dialogue strategy can be chosen for the current turn in the dialogue. PED automatically maintains a belief model by a belief revision process that uses the observed acts in the dialogue. The game tree is evaluated in the context of this belief model. PED is efficient because it uses probabilistic estimates of belief rather than a plain logical belief model.
MARF is a general cross-platform framework with a collection of algorithms for audio (voice, speech, and sound) and natural language text analysis and recognition, along with sample applications (identification, NLP, etc.) of its use. MARF can run distributed over the network (CORBA, Java RMI, and Java XML-RPC Web Services) and may act as a library in applications or be used as a source for learning and extension.
SIOX4Java (Simple Interactive Object Extraction) is a Java SDK that provides a generic segmentation engine for extracting the foreground from still images with little user interaction. The underlying method (which has also been integrated into GIMP) is noise and motion blur robust and can easily be adapted for the segmentation of objects in videos. The SDK also contains an experimental feature called the "Detail Refinement Brush", which enables the removal of spill colors and manual refinement of highly detailed textures.
Mimas Toolkit is a C++ computer vision toolkit. It is easy to use and includes tools for edge detection, corner detection, various filters, optic flow, tracking, blob analysis, Web cam tools for real-time applications, and much more. It also includes many implementations of traditional algorithms such as Canny. It was developed for GNU/Linux but as the GUI is largely separate, porting to other platforms should be straightforward.
VXL is a set of portable C++ libraries designed for computer vision research and implementation. Numerics, imaging, and geometry are provided by stand-alone core libraries, with easy to use APIs and sophisticated processing algorithms. Other libraries provide stereo, video, structure from motion, probability modeling, GUI design, classification, robust estimation, feature tracking, topology, 3d imaging, and much more. It is written and used by an international team from academia and industry.