Skip to main content
Ctrl+K
Tao Toolkit - Home

Tao Toolkit

Tao Toolkit - Home

Tao Toolkit

Table of Contents

Introduction

  • Overview
  • TAO Launcher
  • Migration Guides
    • Migrating to TAO 5.5
    • Migrating from TAO 4.0.x to TAO 5.0.0
    • Migrating from TAO 3.x to TAO 4.0
    • Migrating from Legacy TLT to TAO
    • Migrating to the TAO API from TAO 5.3.0
  • Working With the Containers

Getting Started

  • Getting Started
  • Beginners
  • Intermediate Users
  • Advanced Users

Model Zoo

  • Overview

Running TAO in the Cloud

  • Running TAO in the Cloud
  • Running TAO on an AWS VM
  • Running TAO on Google Cloud Platform
  • Running TAO on an Azure VM
  • Running TAO on Google Colab
  • Running TAO on an EKS
  • Running TAO on an AKS

TAO API

  • Overview
  • Setup
  • Deployment
  • Train, Adapt, Optimize (TAO) REST API
  • Remote Client
  • API Reference

AutoML

  • AutoML

CV Fine-tuning

  • Optimizing the Training Pipeline
  • Visualizing Training
  • Data Annotation Format
  • Computer Vision Finetuning
    • PyTorch
      • Metric Learning Recognition
        • Metric Learning Recognition
      • Instance Segmentation
        • Data Input for Instance Segmentation
        • mask2former
        • Mask Auto Labeler
        • Mask Grounding DINO
      • CenterPose
        • CenterPose
      • Character Recognition
        • OCRNet
      • VisualChangeNet
        • Visual ChangeNet-Segmentation
        • Visual ChangeNet-Classification
      • 3D Object Detection
        • PointPillars
      • ReIdentificationNet Transformer
        • ReIdentificationNet Transformer
      • Optical Inspection
        • SiameseOI
      • Pose Classification
        • PoseClassificationNet
      • Object Detection
        • Grounding DINO
        • DINO
        • OCDNet
        • Deformable DETR
      • ReIdentificationNet
        • ReIdentificationNet
      • ActionRecognitionNet
      • BEVFusion
      • Image Classification PyT
      • SegFormer
    • TensorFlow 2.x
      • Image Classification (TF2)
      • EfficientDet (TF2)
    • TensorFlow 1.x [Deprecated]
      • Body Pose Estimation
        • Body Pose Estimation
      • Character Recognition
        • LPRNet
      • Emotion Classification
      • Facial Landmarks Estimation
        • Facial Landmarks Estimation
      • Gaze Estimation
      • Gesture Recognition
      • HeartRate Estimation
      • Instance Segmentation
        • Data Input for Instance Segmentation
        • MaskRCNN
      • Image Classification (TF1)
      • Multitask Image Classification
      • Object Detection
        • DetectNet_v2
        • FasterRCNN
        • YOLOv3
        • YOLOv4
        • YOLOv4-tiny
        • SSD
        • DSSD
        • RetinaNet
        • EfficientDet (TF1)
      • Semantic Segmentation
        • UNET
  • Foundation Models

MLOPS integration

  • TAO MLOPS Integration
  • TAO WandB Integration
  • TAO Clearml Integration

Deploying to Inference SDKs

  • Integrating TAO CV Models with Triton Inference Server
  • TAO Converter
    • TAO Converter with Classification TF1/TF2
    • TAO Converter with Deformable DETR
    • TAO Converter with Detectnet_v2
    • TAO Converter with DSSD
    • TAO Converter with EfficientDet
    • TAO Converter with FasterRCNN
    • TAO Converter with MaskRCNN
    • TAO Converter with Multitask Classification
    • TAO Converter with Retinanet
    • TAO Converter with SSD
    • TAO Converter with UNET
    • TAO Converter with YOLOv3
    • TAO Converter with YOLOv4
    • TAO Converter with YOLOv4-tiny
  • Optimizing and Profiling with TensorRT
    • TRTEXEC with ActionRecognitionNet
    • TRTEXEC with BodyPoseNet
    • TRTEXEC with CenterPose
    • TRTEXEC with Classification TF1/TF2/PyT
    • TRTEXEC with Deformable-DETR
    • TRTEXEC with DetectNet-v2
    • TRTEXEC with DINO
    • TRTEXEC with DSSD
    • TRTEXEC with EfficientDet TF1/TF2
    • TRTEXEC with Facial Landmarks Estimation
    • TRTEXEC with Faster RCNN
    • TRTEXEC with Grounding DINO
    • TRTEXEC with LPRNet
    • TRTEXEC with Metric Learning Recognition
    • TRTEXEC with Mask RCNN
    • TRTEXEC with Multitask Classification
    • TRTEXEC with OCDNet
    • TRTEXEC with OCRNet
    • TRTEXEC with PointPillars
    • TRTEXEC with PoseClassificationNet
    • TRTEXEC with ReIdentificationNet
    • TRTEXEC with ReIdentificationNet Transformer
    • TRTEXEC with RetinaNet
    • TRTEXEC with Segformer
    • TRTEXEC with SiameseOI
    • TRTEXEC with SSD
    • TRTEXEC with UNet
    • TRTEXEC with YOLO_v3
    • TRTEXEC with YOLO_v4
    • TRTEXEC with YOLO_v4_tiny
    • TRTEXEC with VisualChangeNet
    • TRTEXEC with Mask2former
    • TRTEXEC with Mask Grounding DINO
  • Integrating TAO Models into DeepStream
    • Deploying to DeepStream for Classification TF1/TF2/PyTorch
    • Deploying to DeepStream for Multitask Classification
    • Deploying to DeepStream for DetectNet_v2
    • Deploying to DeepStream for Deformable DETR
    • Deploying to DeepStream for DINO
    • Deploying to DeepStream for DSSD
    • Deploying to DeepStream for EfficientDet
    • Deploying to DeepStream for FasterRCNN
    • Deploying to DeepStream for RetinaNet
    • Deploying to DeepStream for SSD
    • Deploying to DeepStream for YOLOv3
    • Deploying to DeepStream for YOLOv4
    • Deploying to DeepStream for YOLOv4-tiny
    • Deploying to DeepStream for MaskRCNN
    • Deploying to Deepstream for UNet
    • Deploying to Deepstream for Segformer
    • Deploying nvOCDR to DeepStream

Deploying with TAO Deploy

  • TAO Deploy Overview
  • TAO Deploy Installation
  • CenterPose with TAO Deploy
  • Classification (PyTorch) with TAO Deploy
  • Classification (TF1) with TAO Deploy
  • Classification (TF2) with TAO Deploy
  • Deformable DETR with TAO Deploy
  • DINO with TAO Deploy
  • Grounding DINO with TAO Deploy
  • DetectNet_v2 with TAO Deploy
  • DSSD with TAO Deploy
  • EfficientDet (TF1) with TAO Deploy
  • EfficientDet (TF2) with TAO Deploy
  • Faster RCNN with TAO Deploy
  • LPRNet with TAO Deploy
  • Mask RCNN with TAO Deploy
  • Mask2former with TAO Deploy
  • MLRecogNet with TAO Deploy
  • Multitask Image Classification with TAO Deploy
  • OCDNet with TAO Deploy
  • RetinaNet with TAO Deploy
  • SSD with TAO Deploy
  • Segformer with TAO Deploy
  • UNet with TAO Deploy
  • YOLOv3 with TAO Deploy
  • YOLOv4 with TAO Deploy
  • YOLOv4-tiny with TAO Deploy
  • OCRNet with TAO Deploy
  • SiameseOI with TAO Deploy
  • VisualChangeNet-Classification with TAO Deploy
  • VisualChangeNet-Segmentation with TAO Deploy
  • Mask Grounding DINO with TAO Deploy

Bring Your Own Model (BYOM)

  • Bring Your Own Model (BYOM)
    • BYOM Converter
    • BYOM Image Classification

Data Services

  • Annotations
  • Offline Data Augmentation
  • Auto-Label
  • Data Analytics

More Information

  • Release Notes
  • Frequently Asked Questions
  • Troubleshooting Guide
  • Tutorial Videos
    • Getting started with NVIDIA TAO
    • Create Custom Multi-Modal Fusion Models
    • Use Visual Prompt for In-Context Segmentation with NVIDIA TAO
    • Estimate and Track Object Poses with the NVIDIA TAO FoundationPose Model
    • Open Vocabulary Object Detection with NVIDIA Grounding-DINO
    • Use Text Prompts for Auto-Labeling with NVIDIA TAO
    • Visualize Model Training with TensorBoard
  • Support Information
  • Acknowledgements
  • Computer...

Computer Vision Finetuning#

TAO is a pre-training, finetuning, and optimization application for computer vision DNNs (deep neural networks). The pipelines in TAO are implemented across the TensorFlow and PyTorch Deep Learning Frameworks. The details of the finetuning pipelines, including hyperparameters and features, are covered in the subsequent sections.

  • PyTorch
    • Metric Learning Recognition
    • Instance Segmentation
    • CenterPose
    • Character Recognition
    • VisualChangeNet
    • 3D Object Detection
    • ReIdentificationNet Transformer
    • Optical Inspection
    • Pose Classification
    • Object Detection
    • ReIdentificationNet
    • ActionRecognitionNet
    • BEVFusion
    • Image Classification PyT
    • SegFormer
  • TensorFlow 2.x
    • Image Classification (TF2)
    • EfficientDet (TF2)
  • TensorFlow 1.x [Deprecated]
    • Body Pose Estimation
    • Character Recognition
    • Emotion Classification
    • Facial Landmarks Estimation
    • Gaze Estimation
    • Gesture Recognition
    • HeartRate Estimation
    • Instance Segmentation
    • Image Classification (TF1)
    • Multitask Image Classification
    • Object Detection
    • Semantic Segmentation

previous

Data Annotation Format

next

PyTorch

NVIDIA NVIDIA

Copyright © 2025, NVIDIA Corporation.

Last updated on Jun 25, 2025.