Logo
3.0

Introduction

  • Overview
    • Pre-trained Models
    • TLT Workflow Overview
  • Requirements and Installation
    • Hardware Requirements
    • Software Requirements
    • Installation Prerequisites
    • Installation
      • Running the Transfer Learning Toolkit
        • Use the examples
      • Downloading the Models
        • Configure the NGC API key
        • Get a list of models
        • Download a model
  • Open Model Architectures
    • Model Requirements
      • Classification
      • Object Detection
        • DetectNet_v2
        • FasterRCNN
        • SSD
        • DSSD
        • YOLOv3
        • YOLOv4
        • RetinaNet
      • Instance Segmentation
        • MaskRCNN
      • Semantic Segmentation
        • UNet
      • Training
      • Deployment
  • Purpose-Built Pretrained Models
    • Computer Vision
      • Training
      • Deployment
        • TrafficCamNet
        • PeopleNet
        • DashCamNet
        • FaceDetect-IR
        • VehicleMakeNet
        • VehicleTypeNet
        • Emotion Recognition (EmotionNet)
        • Gaze Estimation (GazeNet)
        • Heart Rate Estimation (HeartRateNet)
        • Facial Landmark (FPENet)
        • Gesture Recognition (GestureNet)
        • License Plate Detection (LPDNet)
        • License Plate Recognition (LPRNet)
        • FaceDetect
        • PeopleSegNet
    • Conversational AI
  • Optimizations to the Training Pipeline
    • Quantization Aware Training
    • Automatic Mixed Precision
  • TLT Launcher
    • Software Requirements
    • Installing the launcher
    • Running the launcher
      • Handling launched processes
  • Getting Started With TLT
    • 1. Prerequisites
    • 2. Download Jupyter Notebook
      • Open model architecture:
    • 3. Start Jupyter Notebook
    • 4. Train the Model
  • Deepstream - TLT Integration
    • Prerequisites
      • Software Installation
    • Deployment files
    • Sample Application : License Plate Detection and Recognition
      • Download Repository
      • Download Models
      • Convert to TRT Engine
      • Build
      • Update the paths in model confiuration files
      • Run
    • Resources
      • PeopleNet
      • TrafficCamNet
      • DashCamNet
      • FaceDetectIR
  • Integrating TLT Trained Models to Jarvis
  • Migrating to TLT 3.0

Computer Vision

  • Image Classification
    • Preparing the Input Data Structure
    • Creating an Experiment Spec File - Specification File for Classification
      • Model Config
        • BatchNormalization Parameters
        • Activation functions
      • Eval Config
      • Training Config
        • Learning Rate Scheduler
    • Training the model
      • Required Arguments
      • Optional Arguments
    • Evaluating the Model
      • Required Arguments
      • Optional Arguments
    • Running Inference on a Model
      • Required arguments
      • Optional arguments
    • Pruning the Model
      • Required Arguments
      • Optional Arguments
      • Using the Prune Command
    • Re-training the Pruned Model
    • Exporting the model
      • INT8 Mode Overview
        • Required Arguments
        • Optional Arguments
      • FP16/FP32 Model
      • Exporting the Model
        • Required Arguments
        • Optional Arguments
      • INT8 Export Mode Required Arguments
      • INT8 Export Optional Arguments
      • Exporting a Model
    • Deploying to DeepStream
      • Generating an Engine Using tlt-converter
        • Instructions for x86
        • Instructions for Jetson
        • Using the tlt-converter
      • Integrating the model to DeepStream
        • Integrating a Classification Model
  • Object Detection
    • Offline Data Augmentation
      • Configuring the Augmentor
        • Spatial Augmentation Config
        • Color Augmentation Config
        • Dataloader
        • Blur Config
      • Running the Augmentor Tool
    • Data Input for Object Detection
      • KITTI Format
        • Label Files
        • Sequence Mapping File
    • DetectNet_v2
      • Pre-processing the Dataset
        • Configuration File for Dataset Converter
        • Sample Usage of the Dataset Converter Tool
      • Creating a Configuration File
        • Model Config
        • BBox Ground Truth Generator
        • Post-Processor
        • Cost Function
        • Trainer
        • Augmentation Module
        • Configuring the Evaluator
        • Dataloader
        • Specification File for Inference
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Using Inference on the Model
        • Required Parameters
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
        • Using the Prune Command
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Generating an INT8 tensorfile Using the calibration_tensorfile Command
        • Exporting the DetectNet_v2 Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Sample usage for the export sub-task
      • Deploying to Deepstream
        • Generating an Engine Using tlt-converter
        • Label File
        • DeepStream Configuration File
    • FasterRCNN
      • Preparing the Input Data Structure
        • Required Arguments
        • Optional Arguments
      • Creating an experiment spec file - Specification file for FasterRCNN
        • Dataset
        • Data augmentation
        • Model architecture
        • Training configurations
        • Inference configurations
        • Evaluation configurations
      • Training the model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
        • Using a Pretrained Model
        • Re-training a pruned model
        • Resuming an interrupted training
      • Evaluating the model
        • Required Arguments
        • Optional Arguments
        • Evaluation Metrics
        • Two Modes for Evaluation
      • Running inference on the model
        • Required Arguments
        • Optional Arguments
        • Two Modes for Inference
      • Pruning the model
        • Required Arguments
        • Optional Arguments
        • Using the Prune Command
      • Retraining the pruned model
      • Exporting the model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting the Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Exporting a Model
      • Deploying to DeepStream
        • TensorRT Open Source Software (OSS)
        • Generating an Engine Using tlt-converter
        • Integrating the model to DeepStream
    • YOLOv3
      • Creating a Configuration File
        • Training Config
        • Evaluation Config
        • NMS Config
        • Augmentation Config
        • Dataset Config
        • YOLOv3 Config
      • Generate the Anchor Shape
        • Required Arguments
        • Optional Arguments
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Running Inference on a YOLOv3 Model
        • Required Arguments
        • Optional Arguments
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
        • Using the Prune Command
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting the Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Sample Usage
      • Deploying to DeepStream
        • TensorRT Open Source Software (OSS)
        • Generating an Engine Using tlt-converter
        • Integrating the model to DeepStream
        • Label File
        • DeepStream Configuration File
    • YOLOv4
      • Creating a Configuration File
        • Training Config
        • Evaluation Config
        • NMS Config
        • Augmentation Config
        • Dataset Config
        • YOLO4 Config
      • Generate anchor shape
        • Required Arguments
        • Optional Arguments
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Running Inference on a YOLOv4 Model
        • Required Arguments
        • Optional Arguments
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
        • Using the Prune Command
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting the Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Sample usage
      • Deploying to DeepStream
        • TensorRT Open Source Software (OSS)
        • Generating an Engine Using tlt-converter
        • Integrating the model with DeepStream
        • Label File
        • DeepStream Configuration File
    • SSD
      • Preparing the Dataset
      • Creating a Configuration File
        • Training Config
        • Evaluation Config
        • NMS Config
        • Augmentation Config
        • Dataset Config
        • SSD config
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Running Inference on the Model
        • Required Arguments
        • Optional Arguments
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting command
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Exporting a Model
      • Deploying to Deepstream
        • Prerequisites for SSD Model
        • Label File
        • DeepStream Configuration File
    • DSSD
      • Preparing the Dataset
      • Creating a Configuration File
        • Training Config
        • Evaluation Config
        • NMS Config
        • Augmentation Config
        • Dataset Config
        • DSSD Config
      • Training the Model
        • Required Arguments
        • Optional Arguments
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Running Inference on the Model
        • Required Arguments
        • Optional Arguments
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting command
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Exporting a Model
      • Deploying to Deepstream
        • Prerequisites for DSSD Model
        • Label File
        • DeepStream Configuration File
    • RetinaNet
      • Creating a Configuration File
        • Training Config
        • Evaluation Config
        • NMS Config
        • Augmentation Config
        • Dataset Config
        • RetinaNet Config
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Running Inference on a RetinaNet Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Pruning the Model
        • Required Arguments
        • Optional Arguments
        • Using the Prune Command
      • Re-training the Pruned Model
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting the Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Sample usage
      • Deploying to DeepStream
        • TensorRT Open Source Software (OSS)
        • Generating an Engine Using tlt-converter
        • Integrating the model to DeepStream
        • Label File
        • DeepStream Configuration File
  • Instance Segmentation
    • Data Input for Instance Segmentation
      • COCO format for Instance Segmentation
    • MaskRCNN
      • Creating a Configuration File
        • MaskRCNN Config
        • Data Config
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Running Inference on the Model
        • Required Arguments
        • Optional Arguments
    • Exporting the Model
      • INT8 Mode Overview
      • FP16/FP32 Model
      • Exporting the MaskRCNN Model
        • Required Arguments
        • Optional Arguments
      • INT8 Export Mode Required Arguments
      • INT8 Export Optional Arguments
      • Sample usage
    • Deploying to DeepStream
      • TensorRT Open Source Software (OSS)
        • TensorRT OSS on x86
        • TensorRT OSS on Jetson (ARM64)
      • Generating an Engine Using tlt-converter
        • Instructions for x86
        • Instructions for Jetson
        • Using the tlt-converter
      • Integrating the model with DeepStream
      • Integrating a MaskRCNN Model
        • Label File
        • DeepStream Configuration File
  • Semantic Segmentation
    • Data Input for Semantic Segmentation
    • UNET
      • Creating a Configuration File
        • Model Config
        • Training
        • Dataset
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Sample Usage
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
      • Using Inference on the Model
        • Required Parameters
      • Exporting the Model
        • INT8 Mode Overview
        • FP16/FP32 Model
        • Exporting the UNet Model
        • INT8 Export Mode Required Arguments
        • INT8 Export Optional Arguments
        • Sample Usage for the Export Subtask
      • Deploying to Deepstream
        • Generating an Engine Using tlt-converter
        • Label File
        • DeepStream Configuration File
  • Labeling Data Format
    • Json Label Data Format
  • Emotion Classification
    • Pre-processing the Dataset
      • Sample Usage of the Dataset Converter Tool
    • Creating an Experiment Specification File
      • Trainer
      • Model
      • Loss
      • Optimizer
      • Dataloader
    • Training the Model
      • Required Arguments
      • Optional Arguments
      • Sample Usage
    • Evaluating the Model
      • Required Arguments
      • Optional Arguments
    • Run Inference on the Model
      • Required Parameters
      • Sample usage for the inference sub-task
      • Exporting the EmotionNet Model
        • Required Arguments
      • Sample usage for the export sub-task
    • Deploying to the TLT CV Inference Pipeline
  • Gaze Estimation
    • Pre-processing the Dataset
      • Sample Usage of the Dataset Converter Tool
    • Creating an Experiment Specification File
      • Trainer/Ealuator
      • Model
      • Loss
      • Optimizer
      • Dataloader
      • Augmentation
    • Training the Model
      • Required Arguments
      • Optional Arguments
      • Sample Usage
    • Evaluating the Model
      • Required Arguments
      • Optional Arguments
    • Run Inference on the Model
      • Required Parameters
      • Sample usage for the inference sub-task
      • Exporting the GazeNet Model
        • Required Arguments
      • Sample usage for the export sub-task
    • Deploying to the TLT CV Inference Pipeline
  • Character Recognition
    • LPRNet
      • Preparing the Dataset
      • Creating an Experiment Spec File
        • lpr_config
        • training_config
        • eval_config
        • augmentation_config
        • dataset_config
      • Training the Model
        • Required Arguments
        • Optional Arguments
      • Evaluating the model
        • Required Arguments
        • Optional Arguments
      • Running Inference on the LPRNet Model
        • Required Arguments
        • Optional Arguments
      • Exporting the Model
        • Required Arguments
        • Optional Arguments
      • Deploying the Model
        • Using tlt-converter
        • Deploying the LPRNet in the DeepStream sample
  • Facial Landmarks Estimation
    • Facial Landmarks
    • Dataset Preparation
      • Configuration File for Dataset Converter
      • Sample Usage of the Dataset Converter Tool
    • Creating an Experiment Spec file
      • Trainer Config
      • Model Config
      • Loss Config
      • Dataloader Config
      • Optimizer Config
      • Complete Sample Experiment Spec File
    • Training the model
      • Sample Usage of the Train tool
    • Evaluating the model
      • Sample Usage of the Evaluate tool
    • Inference of the model
      • Sample Usage of the Inference tool
    • Exporting the model
      • Sample Usage of the Export tool
    • Deploying to the TLT CV Inference Pipeline
  • Gesture Recognition
    • Pre-processing the Dataset
      • Image Format
      • Label Format
      • Dataset Extraction Config
      • Dataset Experiment Config
      • Sample Usage of the Dataset Converter Tool
      • Required Arguments
      • Sample Usage
    • Creating a Configuration File
      • Trainer Config
        • Top Training Config
        • Fine Tuning Config
      • Model Config
      • Evaluator Config
    • Training the Model
      • Required Arguments
      • Sample Usage
    • Evaluating the Model
      • Required Arguments
      • Sample Usage
    • Running Inference on the Model
      • Required Arguments
      • Sample Usage
    • Exporting the Model
      • Required Arguments
      • Optional Arguments
      • Sample Usage
    • Deploying to the TLT CV Inference Pipeline
  • Heart Rate Estimation
    • Data Input for Heart Rate Estimation
    • HeartRateNet Data Format
    • Creating a Configuration File To Generate TFRecords
    • Generating TFRecords
      • Required Arguments
      • Sample Usage
    • Creating a Configuration File to Train and Evaluate Heart Rate Network
      • Dataloader
      • Model
      • Loss
      • Optimizer
    • Training the Model
      • Required Arguments
      • Optional Arguments
      • Sample Usage
    • Evaluating the Model
      • Required Arguments
      • Optional Arguments
    • Running Inference on the Model
      • Required Arguments
      • Optional Arguments
    • Exporting the Model
      • Required Arguments
      • Optional Arguments
  • PeopleNet
    • Training algorithm
    • Intended use case
  • LPDNet
    • Training algorithm
    • Intended use case
  • License Plate Recognition Model (LPRNet)
    • Training algorithm
    • Intended use case
  • DashCamNet
    • Training algorithm
    • Intended use case
  • TrafficCamNet
    • Training algorithm
    • Intended Use Case
  • FaceDetect-IR
    • Training algorithm
    • Intended use case
  • VehicleMakeNet
    • Training Algorithm
    • Intended Use
  • VehicleTypeNet
    • Training Algorithm
    • Intended Use
  • PeopleSegNet
    • Training Algorithm
    • Intended Use

TLT Computer Vision Inference Pipeline

  • Overview
  • Requirements and Installation
    • Hardware Requirements
      • Minimum
      • Recommended
    • Software Requirements
    • Installation Prerequisites
    • Installation
      • Configure the NGC API key
      • Download the TLT CV Inference Pipeline Quick Start
  • TLT CV Inference Pipeline Quick Start Scripts
    • Configuring
    • Initialization
    • Launching the Server and Client Containers
    • Stopping
    • Cleaning
    • Integration with TLT
      • Emotion
      • Face Detect
      • Facial Landmarks
      • Gaze
      • Gesture
      • Heart Rate
  • Running and Building Sample Applications
    • TLT CV Inference Pipelines
    • Performance
    • Running the Emotion Classification Sample
      • Emotion API Usage
    • Running the Face Detection Sample
      • Face Detection API Usage
    • Running the Facial Landmarks Estimation Sample
      • Facial Landmarks API Usage
    • Running the Gaze Estimation Sample
      • Gaze Estimation API Usage
    • Running the Gesture Classification Sample
      • Gesture Classification API Usage
    • Running the Heart Rate Estimation Sample
      • Heart Rate Estimation API Usage
    • Building the Sample Applications

Conversational AI

  • ASR
    • Speech Recognition
      • Downloading Sample Spec Files
        • Required Arguments
      • Preparing the Dataset
      • Creating an Experiment Spec File
        • Training Process Configs
        • Dataset Configs
        • Model Configs
      • Training the Model
        • Required Arguments
        • Optional Arguments
        • Training Procedure
        • Troubleshooting
      • Evaluating the Model
        • Required Arguments
        • Optional Arguments
        • Evaluation Procedure
        • Troubleshooting
      • Fine-Tuning the Model
        • Required Arguments
        • Optional Arguments
        • Fine-Tuning Procedure
        • Troubleshooting
      • Using Inference on a Model
        • Required Arguments
        • Optional Arguments
        • Inference Procedure
        • Troubleshooting
      • Model Export
        • Required Arguments
        • Optional Arguments
        • Export Spec File
  • Natural Language Processing
    • Joint Intent and Slot Classification
      • Downloading Sample Spec files
      • Data Format
      • Dataset Conversion
      • Model Training
        • Required Arguments for Training
        • Optional Arguments
        • Training Procedure
      • Model Fine-tuning
        • Required Arguments for Fine-tuning
        • Optional Arguments
        • Fine-tuning Procedure
      • Model Evaluation
        • Required Arguments for Evaluation
        • Evaluation Procedure
      • Model Inference
        • Required Arguments for Inference
        • Inference Procedure
      • Model Export
        • Required Arguments for Export
      • Model Deployment
    • Punctuation and Capitalization
      • Introduction
      • Downloading Sample Spec Files
        • Download Spec Required Arguments
      • Data Input for Punctuation and Capitalization Model
      • Data Format
      • Pre-processing the Dataset
        • Convert Dataset Required Arguments
        • Convert Dataset Optional Arguments
        • Download and Convert Tatoeba Dataset Required Arguments
        • Optional Arguments
      • Training a Punctuation and Capitalization model
        • Required Arguments for Training
        • Optional Arguments
        • Important Parameters
      • Fine-tuning a Model on a Different Dataset
        • Required Arguments for Fine-tuning
        • Optional Arguments
      • Evaluating a Trained Model
        • Required Arguments for Evaluation
        • Optional Arguments
      • Running Inference using a Trained Model
        • Required Arguments for Inference
        • Optional Arguments
      • Model Export
        • Required Arguments for Export
        • Optional Arguments
    • Question Answering
      • Introduction
      • Downloading Sample Spec files
      • Data Format
      • Dataset Conversion
      • Model Training
        • Required Arguments for Training
        • Optional Arguments
        • Training Procedure
      • Model Fine-tuning
        • Required Arguments for Fine-tuning
        • Optional Arguments
        • Fine-tuning Procedure
      • Model Evaluation
        • Required Arguments for Evaluation
        • Evaluation Procedure
      • Model Inference
        • Required Arguments for Inference
        • Inference Procedure
      • Model Export
        • Required Arguments for Export
      • Model Deployment
    • Text Classification
      • Introduction
      • Downloading Sample Spec files
      • Data Format
      • Dataset Conversion
      • Model Training
        • Required Arguments for Training
        • Optional Arguments
        • Training Procedure
      • Training Suggestions
      • Model Fine-tuning
        • Required Arguments for Fine-tuning
        • Optional Arguments
      • Model Evaluation
        • Required Arguments for Evaluation
      • Model Inference
        • Required Arguments for Inference
      • Model Export
        • Required Arguments for Export
    • Token Classification (Named Entity Recognition)
      • Introduction
      • Downloading Sample Spec Files
        • Download Spec Required Arguments
      • Data Input for Token Classification Model
      • Dataset Conversion
        • Convert Dataset Required Arguments
        • Convert Dataset Optional Arguments
      • Training a Token Classification Model
        • Required Arguments for Training
        • Optional Arguments
        • Important Parameters
      • Fine-tuning a Model on a Different Dataset
        • Required Arguments for Fine-tuning
        • Optional Arguments
      • Evaluating a Trained Model
        • Required Arguments for Evaluation
        • Optional Arguments for Evaluation
      • Running Inference using a Trained Model
        • Required Arguments for Inference
        • Optional Arguments
      • Model Export
        • Required Arguments for Export
        • Optional Arguments for Export

More Information

  • Release Notes
    • Transfer Learning Toolkit V3.0
      • Key Features
      • Contents
      • Software Requirements
      • Hardware Requirements
      • Known Issues
      • Resolved Issues
  • Frequently Asked Questions
    • Model support
    • Pruning
    • Model Export and Deployment
    • Training
  • Troubleshooting Guide
    • NGC
    • TLT Launcher
  • Support Information
Transfer Learning Toolkit
  • Docs »
  • ASR
  • View page source

ASRΒΆ

  • Speech Recognition
    • Downloading Sample Spec Files
    • Preparing the Dataset
    • Creating an Experiment Spec File
    • Training the Model
    • Evaluating the Model
    • Fine-Tuning the Model
    • Using Inference on a Model
    • Model Export
Next Previous

© Copyright 2020, NVIDIA. Last updated on Mar 19, 2021.

Built with Sphinx using a theme provided by Read the Docs.