Tao Generate Image Grounding

NVIDIAOfficialCreated by NVIDIA Corporation

Skill
Workflow
Data and AI
Apache-2.0
Medium risk

From the author

Two-step image grounding pipeline: extracts referring expressions from (image, caption) pairs and grounds them to pixel-space bounding boxes via a VLM. Use when the user wants to ground captions to bboxes, generate phrase-grounded annotations, auto-label images for grounding, or run the image_grounding pipeline. Triggers include 'image grounding', 'phrase grounding', 'ground captions', 'auto-label image grounding', 'image_grounding'.

— the description by NVIDIA Corporation

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The tao-generate-image-grounding skill provides a structured workflow for generating phrase-grounded training data. It utilizes a vision-language model (VLM) to perform a two-step process: first, it cleans captions and extracts referring expressions with character spans; second, it predicts pixel-space bounding boxes for those expressions. The pipeline supports various VLM backends, including Gemini and OpenAI-compatible endpoints like NIM or vLLM, and includes built-in checkpointing to allow for efficient, resumable processing of large image datasets.

What it does

Extracts referring expressions and character spans from image captions.
Generates pixel-space bounding boxes for identified expressions.
Supports multiple VLM backends including Gemini and OpenAI-compatible services.
Provides per-sample checkpointing for resumable, fault-tolerant processing.
Allows for flexible pipeline execution, enabling step-by-step or full-run processing.
Outputs structured JSONL annotations suitable for training grounding models.

When to use it

When you need to create phrase-grounded training datasets from existing image-caption pairs.
When you want to auto-label images with bounding boxes for object detection or grounding tasks.
When you need to validate or refine caption-to-image alignment using a VLM.
When you have a large dataset and require a resumable, automated annotation workflow.

Advantages

Automates labor-intensive manual annotation tasks using powerful VLM teacher models.
Highly configurable with support for various self-hosted or cloud-based VLM endpoints.
Resilient workflow design with built-in checkpointing to prevent data loss during long runs.
Flexible execution allows for inspecting intermediate results before full-scale processing.

Usage examples

“Run the image grounding pipeline on my dataset of images and captions.”: The agent will guide you through setting up the input JSONL, selecting your VLM backend (e.g., Gemini or a local vLLM server), and executing the two-step extraction and grounding process.
“I need to extract referring expressions from my captions but not ground them yet.”: The agent will configure the workflow to run only step 0, producing a JSONL file containing cleaned captions and identified expressions with character spans.
“How do I re-run the grounding step after fixing my configuration?”: The agent will instruct you to set workflow.force_reprocess to true to ignore existing checkpoints and re-run the pipeline from scratch.

Install

Download the skill as a ZIP:
https://mp-staging.autogineer.com/agent-skills/nvidia/tao-generate-image-grounding/download

Unzip it into the cross-client skills folder of your project:

mkdir -p .agents/skills && unzip tao-generate-image-grounding.zip -d .agents/skills

Documentation by NVIDIA Corporation

Image Grounding Pipeline

Standalone install? If this session was not initialized by the TAO skill bank plugin, run the tao-setup skill first (host preflight, credentials, cross-skill discovery).

Turn (image, caption) pairs into per-image grounded annotations: cleaned captions, referring expressions with character spans, and pixel-space bounding boxes for each expression. A single VLM (Gemini or any OpenAI-compatible endpoint) handles both steps.

Purpose

Generate phrase-grounded training data for referring-expression and grounding models. The VLM acts as a "teacher" annotator: Step 0 extracts referring expressions from the caption while looking at the image; Step 1 returns one bbox set per expression for each image.

Pipeline Architecture

Step 0: Expression extraction  → VLM cleans caption, extracts referring expressions + char spans
Step 1: Phrase grounding       → VLM returns pixel bboxes + scores per expression

Steps are individually selectable via workflow.steps. Each step writes a per-sample checkpoint to step_<N>_*/.ckpt/<sample_id>.json and skips already-processed records on re-run. Set workflow.force_reprocess: true to ignore checkpoints and reprocess from scratch.

Instructions

Initial setup

When a user wants to run this pipeline, walk through these steps:

Input JSONL: Ask for the JSONL path. Each line must be one object like {"image_path": "...", "caption": "..."}. image_path can be absolute or relative.
Image root: If any image_path values are relative, set data.image_root to the directory they should resolve from.
API access: Ask the user which VLM endpoint they want to use. Present these five options and act on the choice:
1. Gemini — set vlm.backend: "gemini"; require GOOGLE_API_KEY (env var or vlm.gemini.api_key).
2. NIM (e.g. https://inference-api.nvidia.com/v1) — set vlm.backend: "openai"; collect base_url, model_name, and api_key.
3. TAO inference microservice (self-hosted, OpenAI-compatible). Confirm whether the server is already running:
  - Running — collect base_url, model_name, and (optionally) api_key; set vlm.backend: "openai".
  - Not running — guide the user through the skills/applications/tao-run-inference-service skill, which stands up a local TAO inference microservice with an OpenAI-compatible API. Before promising a specific model, check skills/applications/tao-run-inference-service/references/service.yaml for valid_network_arch_config_basenames. Once the server is up, collect base_url, model_name, and (optionally) api_key; set vlm.backend: "openai".
4. vLLM (self-hosted, OpenAI-compatible). Confirm whether the server is already running:
  - Running — collect base_url, model_name, and (optionally) api_key; set vlm.backend: "openai".
  - Not running — follow references/vllm_server.md to install and launch a vLLM server, then collect base_url, model_name, and (optionally) api_key; set vlm.backend: "openai".
5. Custom (any other OpenAI-compatible endpoint) — set vlm.backend: "openai"; collect base_url, model_name, and (optionally) api_key.
If the user has no endpoint and does not want to set one up, stop and help resolve API access first.
Workflow steps: Choose one of:
- Full pipeline: ["0", "1"]
- Expression extraction only: ["0"]
- Grounding only: ["1"], which requires existing step-0 output at results_dir/step_0_expression_extraction/annotations.jsonl
Resume vs fresh run: By default, the workflow reuses checkpoints and skips completed records. To reprocess everything, set image_grounding.workflow.force_reprocess=true.

Running the pipeline

The pipeline runs inside the TAO Toolkit container via the auto_label CLI:

auto_label generate -e /path/to/spec.yaml \
    results_dir=/results \
    image_grounding.data.input_jsonl=/data/captions.jsonl \
    image_grounding.data.image_root=/data/images \
    image_grounding.vlm.gemini.api_key=$GOOGLE_API_KEY

Generate a default spec: auto_label default_specs results_dir=/results module_name=auto_label, then set autolabel_type: "image_grounding". All fields support Hydra dot-notation overrides on the command line.

See references/configuration.md for the full YAML structure, all parameters, model/endpoint setup, and error patterns.

Recommended pilot workflow

Run on 5-10 images with both steps
Inspect step_0_expression_extraction/annotations.jsonl — are cleaned_caption and expressions[] accurate? Are the right noun phrases captured?
Inspect step_1_grounding/annotations.jsonl — do the bboxes in expressions[].instances[] look right? Are confidence scores reasonable?
If quality is insufficient, switch the VLM to a stronger model (e.g. gemini-2.5-pro) or raise media_resolution/max_output_tokens, then re-run with force_reprocess=true.
Scale to the full dataset once satisfied.

Configuration

Key configuration fields (full reference in references/configuration.md):

Field	Default	Description
`workflow.steps`	`["0","1"]`	Which pipeline steps to execute (`"0"` = expressions, `"1"` = grounding)
`workflow.max_workers`	`4`	Parallel threads per step (watch API rate limits)
`workflow.force_reprocess`	`false`	Ignore per-sample checkpoints and reprocess from scratch
`vlm.backend`	`"gemini"`	`"gemini"` or `"openai"` (OpenAI-compatible endpoint)
`data.input_jsonl`	required	Path to input JSONL with `image_path` + `caption` per line
`data.image_root`	`""`	Optional prefix for resolving relative `image_path` entries

Inputs

A single JSONL file at data.input_jsonl. One JSON object per line:

Field	Required	Description
`image_path`	yes	Absolute path, or relative path resolved against `data.image_root`
`caption`	yes	Free-text caption for the image
`image_id`	no	Stable identifier; auto-derived from the filename if missing
`width`, `height`	no	Image dimensions in pixels; default to `1920×1080` for bbox clamping if missing

Outputs

All outputs go to results_dir/:

step_0_expression_extraction/annotations.jsonl — per-record output enriched with cleaned_caption and expressions[] (each with text, expression_id, char_span, noun_chunk, empty instances[]).
step_1_grounding/annotations.jsonl — same records with expressions[].instances[] filled in (each instance has bbox: [x1,y1,x2,y2] in pixel space, score in [0.0, 1.0], and bbox_id).
results_dir/annotations.jsonl — copy of the last step's output for convenience.
step_<N>_*/.ckpt/<sample_id>.json — per-sample checkpoints used for resume.

Prerequisites

Container: nvcr.io/nvidia/tao/tao-toolkit:7.3.0-pyt
API access: At least one VLM endpoint (Gemini API key or OpenAI-compatible endpoint capable of image input)

Files

ATTRIBUTION.mdAdded by AutoGineer954 B
BENCHMARK.md7.1 KiB
LICENSELicence file added by AutoGineer11.2 KiB
SKILL.md7.9 KiB
config/skillspector-baseline.yaml834 B
evals/evals.json819 B
references/configuration.md9.3 KiB
references/skill_info.yaml1.1 KiB
references/vllm_server.md6.8 KiB
skill-card.md4.5 KiB
skill.oms.sig5.4 KiB

references/vllm_server.md

vLLM Server Setup for Vision-Language Models

Use this reference only when the parent SKILL.md points here for the current task. If this file conflicts with current SKILL.md, skill_info.yaml, schemas, or platform/model skills, the current authoritative source wins.

1. Prerequisites
1. Install vLLM
- Option A — Docker (default, recommended)
- Option B — pip (host install, advanced)
1. Pick a VLM checkpoint
1. Launch the server
- Option A — Docker (default, recommended)
- Option B — pip install (host launch, advanced)
1. Verify the endpoint
1. Wire the server into the pipeline
Common issues

This guide walks the user through standing up a self-hosted vLLM server that exposes an OpenAI-compatible /v1/chat/completions endpoint for vision-language models (VLMs). Once the server is running, point the pipeline at it with vlm.backend: "openai" and the matching base_url / model_name / api_key values.

1. Prerequisites

NVIDIA GPU(s) with enough VRAM for the chosen VLM (≥24 GB for 7-8B-class models, ≥80 GB for 32B+ models; tensor-parallel across multiple GPUs is supported via --tensor-parallel-size).
A recent NVIDIA driver and nvidia-container-toolkit (for the recommended Docker path).
Docker (recommended). Python ≥ 3.10 is only needed for the optional host install path.
For gated HuggingFace repos: a HF_TOKEN with access to the model.

2. Install vLLM

Default: Docker (Option A). Use it unless you have a specific reason to install on the host — Docker pins a known-good CUDA/PyTorch/vLLM combination and avoids local environment drift. The pip path (Option B) is provided for advanced users who need to patch vLLM, debug locally, or run on a system where Docker is not available.

Option A — Docker (default, recommended)

docker pull vllm/vllm-openai:latest

Option B — `pip` (host install, advanced)

pip install --upgrade "vllm>=0.7.0"

3. Pick a VLM checkpoint

vLLM supports many vision models; the pipelines just need an OpenAI-compatible chat-completions endpoint that accepts image inputs. Common picks:

Model	HuggingFace repo	Notes
Qwen3-VL-8B-Instruct	`Qwen/Qwen3-VL-8B-Instruct`	Newer Qwen3 family
Qwen3-VL-235B-A22B-Instruct	`Qwen/Qwen3-VL-235B-A22B-Instruct`	MoE; requires serious hardware

If the chosen repo is gated, accept its license on the HuggingFace web UI first, then make HF_TOKEN available in your session before launching.

4. Launch the server

Use the launcher that matches the install path. Both commands listen on 0.0.0.0:8000 and serve an OpenAI-compatible API at /v1. Prefer Option A (Docker) unless you installed via pip in Section 2.

Option A — Docker (default, recommended)

set -a; source /path/to/.env; set +a   # omit if already exported
docker run --runtime nvidia --gpus all \
    -p 8000:8000 \
    -e HF_TOKEN \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    vllm/vllm-openai:latest \
    --model Qwen/Qwen3-VL-8B-Instruct \
    --served-model-name Qwen/Qwen3-VL-8B-Instruct \
    --tensor-parallel-size 1 \
    --gpu-memory-utilization 0.9 \
    --media-io-kwargs '{"video": {"num_frames": -1, "fps": -1}}'

Option B — `pip` install (host launch, advanced)

set -a; source /path/to/.env; set +a   # omit if already exported

vllm serve Qwen/Qwen3-VL-8B-Instruct \
    --host 0.0.0.0 \
    --port 8000 \
    --dtype bfloat16 \
    --served-model-name Qwen/Qwen3-VL-8B-Instruct \
    --tensor-parallel-size 1 \
    --max-model-len 32768 \
    --gpu-memory-utilization 0.9 \
    --media-io-kwargs '{"video": {"num_frames": -1, "fps": -1}}'

Key flags:

--served-model-name <NAME> — value to use later for vlm.openai.model_name. Defaults to the full HF repo path if omitted.
--tensor-parallel-size <N> — number of GPUs to shard the model across.
--max-model-len <N> — context window; image tokens count against this, so leave headroom for multi-image prompts.
--limit-mm-per-prompt image=<N> — max images per request. Bump it if a prompt sends multiple images.
--gpu-memory-utilization <0..1> — lower (e.g. 0.85) if you hit OOM at load time.

The first launch downloads weights to ~/.cache/huggingface; expect several minutes for 7B+ models.

5. Verify the endpoint

Wait for the log line Uvicorn running on http://0.0.0.0:8000, then sanity-check the server.

List served models:

curl http://localhost:8000/v1/models

Send a minimal vision request:

curl http://localhost:8000/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "Qwen3-VL-8B-Instruct",
        "messages": [{
            "role": "user",
            "content": [
                {"type": "image_url", "image_url": {"url": "https://upload.wikimedia.org/wikipedia/commons/thumb/3/3a/Cat03.jpg/640px-Cat03.jpg"}},
                {"type": "text", "text": "Describe this image in one sentence."}
            ]
        }],
        "max_tokens": 128
    }'

A non-empty choices[0].message.content in the response confirms the server is ready.

6. Wire the server into the pipeline

Once verified, collect these three values from the running server and pass them to the pipeline spec:

base_url — e.g. http://localhost:8000/v1 (no /chat/completions suffix). If vLLM runs on another host, use http://<host_or_ip>:8000/v1 and make sure the port is reachable.
model_name — must match --served-model-name exactly.
api_key — vLLM ignores it but the OpenAI SDK requires a non-null string; use "EMPTY" if no auth is configured.

YAML snippet:

vlm:
  backend: "openai"
  openai:
    base_url: "http://localhost:8000/v1"
    model_name: "Qwen3-VL-8B-Instruct"
    api_key: "EMPTY"
    temperature: 0.3
    max_tokens: 4096
    timeout: 300

Common issues

Symptom	Fix
`CUDA out of memory` on startup	Lower `--max-model-len`, drop `--gpu-memory-utilization` (e.g. `0.85`), pick a smaller model, or raise `--tensor-parallel-size` to spread across more GPUs
`Model architectures ['…'] are not supported`	Upgrade vLLM (`pip install -U vllm`) or use a newer Docker tag — VLM support changes per release
401 / 403 during HuggingFace download	Set `HF_TOKEN` in the launch env and accept the model's license on the HuggingFace web UI
First request hangs for minutes	The model is still warming up — wait for the `Uvicorn running` log line and a successful `GET /v1/models`
`image is too large` / token overflow	Pre-resize images before sending, or raise `--max-model-len`
Empty / truncated responses	Raise `vlm.openai.max_tokens` in the pipeline spec; lower `temperature` for more deterministic output

Requirements and compatibility

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Declares allowed tools, Needs an API key

The author says: Requires docker + nvidia-container-toolkit + at least one VLM endpoint (Gemini API key or OpenAI-compatible).

Security

What our scanner noticed· Medium risk

Large base64-encoded blobs
Instructions that use the network

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Created by

NVIDIA Corporation

Publisher

NVIDIA (Official)

Source

NVIDIA/skills · skills/tao-generate-image-grounding · commit fce3cca

Licence

Apache License 2.0 (redistribution allowed)

LICENSE (added by AutoGineer from LICENSE-APACHE)
ATTRIBUTION.md (written by AutoGineer)

Synced

2026-11-08

AutoGineer is not affiliated with or endorsed by NVIDIA Corporation or NVIDIA.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download tao-generate-image-grounding.zip

From the author

AutoGineer summary

What it does

When to use it

Advantages

Usage examples

Install

Documentation by NVIDIA Corporation

Image Grounding Pipeline

Purpose

Pipeline Architecture

Instructions

Initial setup

Running the pipeline

Recommended pilot workflow

Configuration

Inputs

Outputs

Prerequisites

Files

references/vllm_server.md

vLLM Server Setup for Vision-Language Models

Contents

1. Prerequisites

2. Install vLLM

Option A — Docker (default, recommended)

Option B — pip (host install, advanced)

3. Pick a VLM checkpoint

4. Launch the server

Option A — Docker (default, recommended)

Option B — pip install (host launch, advanced)

5. Verify the endpoint

6. Wire the server into the pipeline

Common issues

Requirements and compatibility

Security

Attribution and licence

Related skills

Pytorch Lightning

Aeon

Umap Learn

Scikit Learn

Shap

Pufferlib

Download ZIP

Option B — `pip` (host install, advanced)

Option B — `pip` install (host launch, advanced)