Traditional artificial neural networks are an interconnected group of nodes (neurons).
The connections between these neurons are modelled as weights and all inputs to the network are modified
based on these weights and summed to produce an outcome.
In contrast, SLIM™ can generate neural networks by making use of just enough information (“See Less”),
which is selected intelligently by our algorithms, at a much faster rate.
By sensing only limited amount of the input data, taken from a camera or any sensor,
classification accuracy of 85% and 70% on MNIST and CIFAR, respectively, benchmark datasets.
Our ‘See Less. Infer More’ models also have a major impact on hardware resource allocation, e.g. low memory devices for IoT.
SLIM™ can be deployed on cloud services, e.g. aws or google and it can also be deployed on memory constraint and battery budgeted edge devices (e.g. mobile phones and FPGAs).
Our way of building models will also enable robots to see (i.e. detection,classification and localisation) in hazy, rainy or foggy environments.