pearlnaturalvision
Pearl Natural Vision (Ver 4.2) – innovative natural, general, multiple, real-time, content-aware 2-3D human-like bio/brain inspired humanoid robot VISION technology, one that can “see” like humans.
Might be interesting for people using Matlab, GIMP, OpenCV, Photoshop, etc.
Related: Automatic & Accurate Extract Extraction, artificial intelligence (AI), deep neural-networks (DNN), multilayer, convolution, computer vision, big data, deep learning, machine learning, generic, autonomous, visual perception, patterns, recognition, detection, saliency, background removal, crowd/complex env’t, invariant, image processing, segmentation, augmented reality, etc.
Also related: psychology, evolution, genetics, adaptation, the Human Brain.
Self made,
Only software,
Written in C++.
Not using any vision libraries.
Highlights:
1) An integration/fusion of Modern Physics, DNN, Neuroscience, Perception, etc., into a Unified, Coherent, General, Accurate and Real Time Bio-like Visual Perception Model.
2) No keys, no params, no settings and no specific visual knowledge, of any kind, are used.
In learning, the results of this visual perception activity (objects/entities/etc…) are placed in the Visual Memory (also a DNN).
After learning phase… naturally the system power multiplies…
Applications:
——————————————-
Robotic humanoid & android vision, smart toys, drones, autonomous driving, guidance, tracking, inspection, face detection, collision prevention, assistance and warning systems, security systems, surveillance, satellite imagery, medical systems, etc.
———-
Email: pearlnaturalvision@gmail.com
Facebook: pearlnaturalvision
2012 2013 2014 2015 2016 2017
Source
wow amazing and that too withthout traning.
Very very intuitive and cool! can i feed it a batch of images for automatic object extraction and will it do it? say a 1000 pictures, will it recognize pictures that have cats in them?
cool