skills.rest
NVIDIA-TAO
NVIDIA-TAOOFFICIAL
Explore all skills by this author →
GitHub

tao-train-rtdetr

Automate RT-DETR model training, evaluation, and TensorRT engine deployment.

83|20|Updated Apr 16, 2026
Search Tags:#deployment#evaluation#training#object detection#real-time detection
One-click install
npx skills add https://github.com/NVIDIA-TAO/tao-skill-bank --skill tao-train-rtdetr
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: tao-train-rtdetr
Source: https://github.com/NVIDIA-TAO/tao-skill-bank/tree/main/skills/models/tao-train-rtdetr
Command: npx skills add https://github.com/NVIDIA-TAO/tao-skill-bank --skill tao-train-rtdetr

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires docker, nvidia-container-toolkit, torch, torchvision, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

The tao-train-rtdetr skill simplifies the process of training, evaluating, and deploying Real-Time Detection Transformer (RT-DETR) models, making object detection and classification tasks more efficient and accurate.

Core Features & Use Cases

  • Training and Evaluation: Automate the training and evaluation of RT-DETR models with ease.
  • Deployment: Generate TensorRT engines for efficient real-time inference.
  • Use Case: A data scientist or AI developer may use this skill to train an RT-DETR model for real-time object detection in videos or images, which can then be deployed for inference using TensorRT engines for faster performance.

Quick Start

To train an RT-DETR model with a custom dataset, run:

/scripts/train.sh train_pretrained_model_path="path/to/your/pretrained/model" train_data_sources="path/to/your/train/data" eval_data_sources="path/to/your/evaluation/data" model_backbone="resnet_50"

Frequently Asked Questions about tao-train-rtdetr

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I train an RT-DETR model for real-time object detection?▼

To train an RT-DETR model for real-time object detection, run the provided training script with your custom dataset paths and a specified backbone like resnet_50. This automates the training process using the PyT CLI.

What is RT-DETR model quantization and distillation used for in object detection?▼

Quantization and distillation for RT-DETR object detection models are used for deployment optimization. They help generate efficient TensorRT engines that enable faster real-time inference on target hardware.

Do I need Docker and NVIDIA container toolkit to deploy RT-DETR models?▼

Yes, you need Docker and the NVIDIA container toolkit to deploy RT-DETR models. This Skill requires these dependencies to manage the containerized environment for training, evaluation, and TensorRT engine generation.

Can I use PyTorch and Torchvision to evaluate real-time detection transformers?▼

Yes, you can use PyTorch and Torchvision to evaluate real-time detection transformers. The Skill utilizes the RT-DETR PyT CLI built on these frameworks to automate the evaluation of 2D object detection models.

How do I export a trained object detection model to a TensorRT engine?▼

To export a trained object detection model to a TensorRT engine, use the export and deployment features provided by the RT-DETR PyT CLI. This process supports distillation and quantization for optimized real-time inference.

Related Skills

Explore More Software Engineering Skills→
NVIDIA-TAONVIDIA-TAO
83

tao-train-grounding-dino

Automate Grounding DINO training, evaluation, and ONNX export with PyTorch.

Official
Advanced
NVIDIANVIDIA
13.2k

trt-onnx-quickstart

Convert ONNX models to TensorRT engines and verify numerical parity.

Official
Advanced
olibartfastolibartfast

deepstream-import-vision-model

Automate HuggingFace or NVIDIA NGC object detection model integration into NVIDIA DeepStream pipelines.

Community
Advanced
AxGordAxGord
5

domain-yolo

Compare YOLO object detection models by speed, accuracy, and memory usage.

Community
Intermediate