# Tao mine aoi images

> Tao mine aoi images is an open-source agent skill for design and media. Mines look-alike source images for VCN augmentation via DEFT embedding. 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-mine-aoi-images
- **Repository:** https://github.com/nvidia/skills
- **GitHub stars:** 4,037
- **License:** Apache-2.0
- **Skill name:** tao-mine-aoi-images
- **Collection:** https://github.com/NVIDIA/skills
- **Categories:** [design and media](https://ai-powered.com/categories/design-and-media), [AI models and LLM tooling](https://ai-powered.com/categories/ai-and-llm)
- **Listing updated:** 2026-11-18

## About

This skill runs a fixed embed-then-mine pipeline for VCN AOI iterations. It embeds a parquet of weak target images and a candidate source pool with the same encoder, then finds the nearest source neighbours to each target and writes a deduplicated parquet of mined paths for the next training round.

The work runs through three containerized steps against a pinned TAO data-services image, requiring Docker, the NVIDIA container toolkit and a CUDA GPU. Iterative runs can also apply a bundled Python script that keeps a history ledger so samples selected in earlier iterations are not picked again, and it writes a summary of the candidate and selection counts.

Inputs are parquet files with a filepath column, and an optional label column enables label-aware filtering. CSV source pools must first be converted to parquet, and the skill expects a shell and Python environment with access to the container image.

## When your agent uses it

> Runs the DEFT embed-then-mine workflow for VCN AOI iterations — embeds the gap-analysis target parquet, embeds a source pool, and mines nearest-neighbour source images for downstream augmentation. Use as the immediate next step after `tao-route-visual-changenet-samples` when expanding a real-image augmentation queue from the mining subset.

## What it can access

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

- [Runs shell and Python scripts](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-mine-aoi-images)
- [Asks for: Read, Bash](https://github.com/NVIDIA/skills/blob/HEAD/skills/tao-mine-aoi-images/SKILL.md)
- [No network calls found in its scripts](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-mine-aoi-images)
- [License: Apache-2.0](https://github.com/NVIDIA/skills/tree/HEAD/skills/tao-mine-aoi-images)

## Features

- **Target and pool embedding:** Encodes target and source images with one shared encoder configuration so the embeddings are comparable.
- **Nearest-neighbour mining:** Finds the closest source images to each target under a chosen distance metric and deduplicates the results.
- **Label-aware filtering:** Optionally restricts matches by label when both embedding files carry a label column.
- **History-aware selection:** Uses a ledger script to exclude samples already selected in earlier iterations of a run.
- **Run reporting:** Saves a mining summary with query, neighbour and duplicate counts alongside the output parquet.

## Good for

- Find real source images similar to weak target images for augmentation
- Expand a real-image augmentation queue from a gap-analysis output
- Run an iterative mining round that avoids repeating earlier selections
- Compare mining results across different top-n or distance settings

## Install

### Claude Code

Copy the skill folder to `~/.claude/skills/tao-mine-aoi-images` 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-mine-aoi-images &&
cp -R "$tmp/skills/tao-mine-aoi-images" ~/.claude/skills/tao-mine-aoi-images &&
rm -rf "$tmp"
```

For one project only, put the folder in `.claude/skills/tao-mine-aoi-images` inside it instead.

### Codex

Copy the skill folder to `~/.agents/skills/tao-mine-aoi-images` 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-mine-aoi-images &&
cp -R "$tmp/skills/tao-mine-aoi-images" ~/.agents/skills/tao-mine-aoi-images &&
rm -rf "$tmp"
```

For one repository only, put the folder in `.agents/skills/tao-mine-aoi-images` inside it instead.

### Cursor

Copy the skill folder to `~/.cursor/skills/tao-mine-aoi-images` 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-mine-aoi-images &&
cp -R "$tmp/skills/tao-mine-aoi-images" ~/.cursor/skills/tao-mine-aoi-images &&
rm -rf "$tmp"
```

For one project only, put the folder in `.cursor/skills/tao-mine-aoi-images` 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-mine-aoi-images
```

Or copy the folder to `~/.gemini/skills/tao-mine-aoi-images` yourself.

### VS Code

For GitHub Copilot in VS Code, copy the skill folder to `~/.copilot/skills/tao-mine-aoi-images`:

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

For one project only, put the folder in `.github/skills/tao-mine-aoi-images` inside it instead.

### Other agents

Get the folder:

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

Then copy `skills/skills/tao-mine-aoi-images` into your agent's skills folder as `tao-mine-aoi-images`. Agents that support SKILL.md read one folder per skill.

## FAQ

### What hardware and software does it need?

Docker with the NVIDIA container toolkit and a CUDA GPU, since both the encoder and the nearest-neighbour search need CUDA. It pulls a pinned TAO data-services container image.

### What inputs does it expect?

A target parquet with a filepath column, typically the output of the gap-analysis or routing step, and a source pool parquet with a filepath column. CSV pools must be converted to parquet first.

### Does it change my training data?

It writes new parquet files and summaries into a timestamped results folder. It does not modify the source images or the input parquets.

### Which encoder does it use by default?

The default embedding spec uses SigLIP with the google/siglip-base-patch16-224 model and a batch size of 64, and the same spec must be used for both embedding steps.

### How do I install Tao mine aoi images in Claude Code?

Copy the skill folder from GitHub to ~/.claude/skills/tao-mine-aoi-images to use it in all your projects, or to .claude/skills/tao-mine-aoi-images in one project.

### Does Tao mine aoi images run scripts?

Yes. It runs shell and Python scripts. No network calls were found in its 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 mine aoi images?

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 mine aoi images?

The closest alternatives to Tao mine aoi images are Tao mine od images, Tao generate image embeddings, Tao train image classification, Tao train nvpanoptix3d and Tao train rtdetr.

## Where this comes from

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

## Alternatives

- [Tao mine od images](https://ai-powered.com/skills/tao-mine-od-images): Tao mine od images is an open-source agent skill for design and media.
- [Tao generate image embeddings](https://ai-powered.com/skills/tao-generate-image-embeddings): Tao generate image embeddings is an open-source agent skill for design and media.
- [Tao train image classification](https://ai-powered.com/skills/tao-train-image-classification): Tao train image classification is an open-source agent skill for design and media.
- [Tao train nvpanoptix3d](https://ai-powered.com/skills/tao-train-nvpanoptix3d): Tao train nvpanoptix3d is an open-source agent skill for design and media.
- [Tao train rtdetr](https://ai-powered.com/skills/tao-train-rtdetr): Tao train rtdetr is an open-source agent skill for ai models and llm tooling.
- [Tao mine nearest neighbors](https://ai-powered.com/skills/tao-mine-nearest-neighbors): Tao mine nearest neighbors is an open-source agent skill for analytics and data.
- [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 train dinov3](https://ai-powered.com/skills/tao-train-dinov3): Tao train dinov3 is an open-source agent skill for ai models and llm tooling.

---
Source: https://ai-powered.com/skills/tao-mine-aoi-images
