.
FRDCSA | minor codebases | Object Recognition
Homepage

[Project image]
Object Recognition

Jump to: Project Description | Parent Description

Project Description

Updating the inventory of the pantry and the rooms using barcodes has proven infeasible, since often times things get really moved around much quicker than the information can be updated. So we are working on for pantry/inventory a combination of RFID in addition to image-based object recognition. The proliferation of deep learning has accelerated the capabilities of such systems. We intend to locate, either through GitHub or provided by some colleagues, the appropriate tools to update the inventory from a cell phone or surveillance camera. The strategy used by the image compontent (i.e. not the RFID) is this, when an object is recognizeds, its absolute location and relative location is recorded along with the recognition confidence, and asserted into the KB's relevant temporal contexts. When a particular item is sought, it checks the last known location and if present the object's assigned default location, otherwising marking the object as misplaced or via user saying directly so. Image models will probably enable a web-based image-based inventory manager, similar to the 3d one that was generously written for the project using Blender.


This page is part of the FWeb package.
Last updated Sat Oct 26 16:43:02 EDT 2019 .