DIGITS is an interactive deep neural network training application for developers to rapidly train highly accurate neural networks for image classification, segmentation and object detection.
Trained models can be deployed to the cloud, PC, embedded or automotive GPU platforms, with TensorRT inference engine that maximizes throughput and efficiency.
Highlights from this release:
- NVIDIA DIGITS 5 introduces image segmentation workflow for partitioning images into regions of interest, and Model Store for downloading pre-trained models that can decrease training time and improve model accuracy.
- NVIDIA TensorRT is a high performance deep learning inference engine for production deployment of applications such as image classification, segmentation, and object detection that delivers up to 14x more images/sec than CPU-only inference.