By Amulya Vishwanath
Announcing Transfer Learning Toolkit General Availability
Warehouses, cities, retail and factory floors are going through massive transformations to support the need for an intelligent and sustainable environment powered by artificial intelligence.
Building an Intelligent Video Analytics (IVA) pipeline by training deep neural networks from scratch is a time consuming process and requires careful consideration while designing and deploying over time. With Transfer Learning Toolkit (TLT), you can train and scale quickly and efficiently.
The term “transfer learning” implies that you can extract learned features from an existing neural network and transfer these learned features by transferring weight from an existing neural network. Developers and data scientists can benefit from using TLT to reduce the time to train, re-adapt and fine-tune DNNs. The toolkit comes pre-packaged with several image classification and detection models and includes networks which have been trained on publicly available datasets. TLT makes multi-gpu and cross-resolution training easy to use. TLT also assists with pruning networks to tightly pack complex applications, delivering high throughput and stream density. When used in conjunction with DeepStream SDK, this becomes an end-to-end versatile deep learning solution for IVA.
The TLT pre-trained models are easily accessible from NVIDIA NGC. Object detection frameworks include Faster RCNN, SSD and DetectNet_v2 (detection technology developed at NVIDIA). Find the full list of supported models here.
Each model is trained for highest accuracy levels and optimized for inference on NVIDIA GPUs (Pascal, Volta, and Turing GPUs) with TensorRT and DeepStream SDK. All the detection models support both RGB and monochrome images as input.
We are here to help you solve the greatest challenges in intelligent video analytics. Check us out and let us know what you think on DevTalk forum.
Getting started with Transfer Learning Toolkit:
- Download Transfer Learning Toolkit here.
- Read the step by step tutorial for building and deploying accurate deep learning models for intelligent image and video analytics on Medium
- Register now and learn how to create and manage real-time object detection applications for disaster response using TLT in this developer webinar.