Responding to disasters and rapidly changing humanitarian crisis events requires developers and engineers to create or update applications to meet requirements quickly and accurately. Reducing the complexity when fine tuning deep neural networks, then easily creating an efficient model for inference purposes, can be key to deploying applications to respond to these events as quickly as possible.

To help with the fine tuning and deployment workflow NVIDIA has created the Transfer Learning Toolkit (TLT) – which in conjunction with DeepStream SDK offers a complete end-to-end solution for real time object detection in video streams. Transfer Learning Toolkit has a unique workflow that allows developers to download the latest state-of-the-art pre-trained models from NVIDIA NGC, fine-tune and adapt these models using custom datasets, as well as perform optimizations such as model pruning.

• Learn data labeling, augmentation and pre-processing

• Use the toolkit to quickly build fast and accurate object detection models for disaster response requirements

• Export these models using TensorRT and perform real-time inference on NVIDIA Jetson Xavier edge devices with DeepStream SDK

Content

content goes here