Welcome to TLT-Trainer Docs

TLT-Trainer can be used to automatically train models of different architectures on the same data using Nvidia Transfer Learning Toolkit.

The Transfer Learning Toolkit - Trainer for Intelligent Video Analytics, This Documentation provides:

  1. General overview and capabilities of the Transfer Learning Toolkit.

  2. Instruction on using TLT-Trainer container to start training models with different architectures and backbones.

  3. Explanation about API, code structure, and overall workflow for development and debugging.

Overview

This software is used to train computer vision and deep learning models for streaming analytics use cases.

The TLT-Trainer downloads pre-trained weight for each backbone from NGC or can use custom models, each base model is trained on Open Images Dataset. Pre-trained weights provide a great starting point for applying transfer learning on your own dataset.

Currently, the following applications are supported:

  • Object Detection

  • Classification

Under object detection the following meta architectures are supported:

  • DetectNet_v2

  • SSD

  • DSSD

  • YOLOv3

  • FasterRCNN

  • RetinaNet

Prerequisites

Get an NGC API key

  1. Go to NGC and click the Transfer Learning Toolkit container in the Catalog tab. This message is displayed, Sign in to access the PULL feature of this repository.

  2. Enter your email address and click Next or click Create an Account.

  3. Choose your organization when prompted for Organization/Team.

  4. Click Sign In.

  5. Select the Containers tab on the left navigation pane and click the Transfer Learning Toolkit tile.

In the next section, you will learn about architecture specific models and their requirements.