Keywords ASICs, neural networks, OCR, parallel architectures, silicon compilers, surface-mounted device assembly, vision systems
Start Date: 01 January 1991 / Duration: 37 months / Status: ongoing
[ contact / participants ]
a hardware and software environment for developing neural network applications and its validation on selected industrial applications
The programming environment consists of:
Industrial applications cover two areas: optical character recognition (OCR) of printed documents and industrial vision for quality control. The industrial vision work focuses on improving both SMD (surface mounted device) assembly techniques and video-based grading systems for fruit.
The specification phase resulted in a full definition of the VML language, which is now used in other ESPRIT neurocomputing projects.
Demonstrators have been implemented and used as a vehicle to validate the hardware and software concepts. The complete system chain, starting from the specification of a neural network application with the graphic tools developed in the project, its generation into N, compilation into VML and execution on the demonstrators has now been completed.
Work on the silicon compiler has progressed; integration of all parts is underway.
A full OCR system has been released and other applications are being ported onto dedicated hardware prior to integration into the industrial systems.
A software environment, MIMENICE, for developing neural applications is being marketed as well as a software system for OCR integrating neural techniques.
SYNAPSE-1, a hardware system for large-scale neural network applications, is now being successfully marketed.
160 Blvd. de Valmy
F-92704 Colombes Cedex
tel: +33 1 4130 3597
fax: +33 1 4130 4918
telex: 616780 THOM F
Computer Technology Institute [GR]
Mimetics SA [F]
Philips [F, NL]
SGS-Thomson Microelectronics SRL [I]
Siap Sistemi Spa [I]
Siemens AG [D]
University College London [UK]
GALATEA - 5293, December 1993
please address enquiries to the ESPRIT Information Desk
html version of synopsis by Nick Cook