# Tao train mask grounding dino

> Tao train mask grounding dino is an open-source agent skill for computer vision. Trains and runs NVIDIA TAO Mask Grounding DINO for text-prompted instance segmentation. It is available on Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose.

- **Kind:** Skill
- **Pricing:** Open source
- **Platforms:** Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
- **Website:** https://github.com/NVIDIA/skills/tree/main/skills/tao-train-mask-grounding-dino
- **Repository:** https://github.com/nvidia/skills
- **GitHub stars:** 4,037
- **License:** Apache-2.0
- **Skill name:** tao-train-mask-grounding-dino
- **Collection:** https://github.com/NVIDIA/skills
- **Categories:** computer vision
- **Listing updated:** 2026-11-18

## About

This skill guides an agent through the lifecycle of an NVIDIA TAO Mask Grounding DINO model: training on segmentation datasets in ODVG, COCO or raw COCO formats, evaluation, inference with text captions, export and quantization. It builds on Grounding DINO by adding a mask prediction head, and it lays out the spec overrides needed for each action, including the dataset paths that must be set for every run.

The skill also covers operational details: default hyperparameters such as the number of region queries and loss coefficients, multi-GPU settings, and the expected validation metric for training. It explains how checkpoints are chosen from a parent job, and it points to a separate reference for TensorRT engine generation and deployment. Training is handled through an AutoML-aware policy that the agent resolves from the model's metadata before launching.

Running it requires Docker with the NVIDIA container toolkit and a GPU with substantial memory, with 24GB or more recommended. The skill is licensed under Apache-2.0 and declares Read and Bash as its allowed tools.

## When your agent uses it

> Mask Grounding DINO for grounded instance segmentation. Extends Grounding DINO with a mask-prediction head for

## What it can access

From a static check of the files on GitHub; nothing was run.

- [Instructions only, no scripts](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-train-mask-grounding-dino)
- [Asks for: Read, Bash](https://github.com/NVIDIA/skills/blob/HEAD/skills/tao-train-mask-grounding-dino/SKILL.md)
- [License: Apache-2.0](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-train-mask-grounding-dino)

## Features

- **Model training workflow:** Covers launching training with ODVG or COCO data, with spec defaults for epochs, learning rate and queries.
- **Mask head configuration:** Explains mask prediction settings, loss types and region query counts that distinguish this model from Grounding DINO.
- **Evaluation and inference:** Describes evaluation thresholds for mask matching and inference driven by text prompt captions.
- **Export and quantization:** Lists the steps and data sources needed to export and quantize a trained checkpoint.
- **Checkpoint selection:** Explains how to pick a specific epoch checkpoint from a parent job rather than the latest symlink.
- **Multi-GPU launch settings:** Documents GPU count, node count and DDP or FSDP strategy options for Lightning-managed runs.

## Good for

- Train a text-prompted segmentation model on my labeled dataset
- Evaluate a trained Mask Grounding DINO checkpoint on a test set
- Run open-vocabulary segmentation on images using custom class captions
- Export and quantize a trained segmentation model

## Install

### Claude Code

Copy the skill folder to `~/.claude/skills/tao-train-mask-grounding-dino` to use it in all your projects:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.claude/skills &&
rm -rf ~/.claude/skills/tao-train-mask-grounding-dino &&
cp -R "$tmp/skills/tao-train-mask-grounding-dino" ~/.claude/skills/tao-train-mask-grounding-dino &&
rm -rf "$tmp"
```

For one project only, put the folder in `.claude/skills/tao-train-mask-grounding-dino` inside it instead.

### Codex

Copy the skill folder to `~/.agents/skills/tao-train-mask-grounding-dino` to use it in every repository:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.agents/skills &&
rm -rf ~/.agents/skills/tao-train-mask-grounding-dino &&
cp -R "$tmp/skills/tao-train-mask-grounding-dino" ~/.agents/skills/tao-train-mask-grounding-dino &&
rm -rf "$tmp"
```

For one repository only, put the folder in `.agents/skills/tao-train-mask-grounding-dino` inside it instead.

### Cursor

Copy the skill folder to `~/.cursor/skills/tao-train-mask-grounding-dino` to use it in every project:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.cursor/skills &&
rm -rf ~/.cursor/skills/tao-train-mask-grounding-dino &&
cp -R "$tmp/skills/tao-train-mask-grounding-dino" ~/.cursor/skills/tao-train-mask-grounding-dino &&
rm -rf "$tmp"
```

For one project only, put the folder in `.cursor/skills/tao-train-mask-grounding-dino` inside it instead.

### Gemini CLI

Run in a terminal (add `--scope workspace` to install it for the current workspace only):

```sh
gemini skills install https://github.com/NVIDIA/skills.git --path skills/tao-train-mask-grounding-dino
```

Or copy the folder to `~/.gemini/skills/tao-train-mask-grounding-dino` yourself.

### VS Code

For GitHub Copilot in VS Code, copy the skill folder to `~/.copilot/skills/tao-train-mask-grounding-dino`:

```sh
tmp=$(mktemp -d) &&
git clone --depth 1 https://github.com/NVIDIA/skills.git "$tmp" &&
mkdir -p ~/.copilot/skills &&
rm -rf ~/.copilot/skills/tao-train-mask-grounding-dino &&
cp -R "$tmp/skills/tao-train-mask-grounding-dino" ~/.copilot/skills/tao-train-mask-grounding-dino &&
rm -rf "$tmp"
```

For one project only, put the folder in `.github/skills/tao-train-mask-grounding-dino` inside it instead.

### Other agents

Get the folder:

```sh
git clone --depth 1 https://github.com/NVIDIA/skills.git
```

Then copy `skills/skills/tao-train-mask-grounding-dino` into your agent's skills folder as `tao-train-mask-grounding-dino`. Agents that support SKILL.md read one folder per skill.

## FAQ

### What does this skill need to run?

It requires Docker with nvidia-container-toolkit, and the compatibility note recommends a GPU with 24GB or more of memory.

### Which dataset formats does training accept?

Training accepts ODVG, COCO and COCO raw formats, with validation using COCO-format annotations.

### Which tools does the skill allow the agent to use?

It declares Read and Bash as its allowed tools.

### Does the skill include scripts?

No scripts are bundled with it; it relies on the TAO CLI and reference files for specs.

### How do I install Tao train mask grounding dino in Claude Code?

Copy the skill folder from GitHub to ~/.claude/skills/tao-train-mask-grounding-dino to use it in all your projects, or to .claude/skills/tao-train-mask-grounding-dino in one project.

### Does Tao train mask grounding dino run scripts?

No. It is instructions only, with no scripts. This comes from a static check of the files on GitHub; nothing was run. Read the files before you install it.

### Which AI agents can use Tao train mask grounding dino?

Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf and Goose. Each reads skills in the SKILL.md format from its own folder; the install steps on this page list them.

### What are the best alternatives to Tao train mask grounding dino?

The closest alternatives to Tao train mask grounding dino are Tao train reid, Tao run deft aoi, Tao train foundation stereo, Tao run deft object detection and Tao train oneformer.

## Where this comes from

- Description: from the skill's own SKILL.md.

## Alternatives

- [Tao train reid](https://ai-powered.com/skills/tao-train-reid): Tao train reid is an open-source agent skill for computer vision.
- [Tao run deft aoi](https://ai-powered.com/skills/tao-run-deft-aoi): Tao run deft aoi is an open-source agent skill for computer vision.
- [Tao train foundation stereo](https://ai-powered.com/skills/tao-train-foundation-stereo): Tao train foundation stereo is an open-source agent skill for design and media.
- [Tao run deft object detection](https://ai-powered.com/skills/tao-run-deft-object-detection): Tao run deft object detection is an open-source agent skill for computer vision.
- [Tao train oneformer](https://ai-powered.com/skills/tao-train-oneformer): Tao train oneformer is an open-source agent skill for computer vision.
- [Tao analyze changenet rca](https://ai-powered.com/skills/tao-analyze-changenet-rca): Tao analyze changenet rca is an open-source agent skill for computer vision.
- [Tao analyze gaps visual changenet](https://ai-powered.com/skills/tao-analyze-gaps-visual-changenet): Tao analyze gaps visual changenet is an open-source agent skill for computer vision.
- [Tao train ocdnet](https://ai-powered.com/skills/tao-train-ocdnet): Tao train ocdnet is an open-source agent skill for computer vision.

---
Source: https://ai-powered.com/skills/tao-train-mask-grounding-dino
