Neurocrafts provides tools to support embedded vision companies in accelerating their productization of computer vision models built on cutting edge technologies using neural networks. The Neurocrafts team excels at providing services based on its products for image and video analytics. Neurocrafts’ tool Model Cleaver automates a complex phase of embedded vision system building. The tool addresses edge computing challenges and converts complex CNN models in a completely automated fashion, enabling them to port on to target edge devices. The following features are supported: knowledge distillation; model pruning and calibration; model quantization; porting the models to resource-constrained hardware; hybrid model generation; and low-bit representations of models with acceptable accuracy. The tool helps in cutting development time down by 70% training time, at 25% of direct cost. Neurocrafts is continuing its research to come up with more sophisticated tools to make embedded vision an easy to adopt technology for small to larger companies.
.