skills.rest
NVIDIA
NVIDIAOFFICIAL
Explore all skills by this author →
WebsiteGitHub

physical-ai-image-attribute-augmentation

Orchestrates image attribute augmentation and auto-labeling workflows on NVIDIA OSMO.

3.2k|370|Updated Feb 25, 2026
Search Tags:#NIM#osmo#auto-labeling#physical ai#person re-identification#Workflow orchestration#image augmentation
One-click install
npx skills add https://github.com/NVIDIA/skills --skill physical-ai-image-attribute-augmentation
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: physical-ai-image-attribute-augmentation
Source: https://github.com/NVIDIA/skills/tree/main/skills/physical-ai-image-attribute-augmentation
Command: npx skills add https://github.com/NVIDIA/skills --skill physical-ai-image-attribute-augmentation

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Running image attribute augmentation and person-attribute captioning pipelines on NVIDIA OSMO involves error-prone steps: choosing the right workflow, deriving storage URLs, verifying NIM inference endpoints, validating credentials, and monitoring long-running jobs. This Skill guides an agent through that entire gated sequence so runs are reproducible and failures surface early.

Core Features & Use Cases

  • Flow Selection and Orchestration: Maps user intent to one of three OSMO workflows (e2e, augmentation-only, auto_labeling-only) and submits them with validated interpolation values.
  • Preflight and Endpoint Gating: Runs credential preflight checks, verifies Image Edit, VLM, and LLM NIM endpoint health, and stops with clear remediation when prerequisites fail.
  • Monitoring and Output Retrieval: Provides OSMO commands for workflow status, per-task logs, and downloading augmented images, captions, and dataset JSON.
  • Use Case: A robotics engineer with person-crop images in S3 asks the agent to generate clothing variations and attribute captions; the Skill selects the e2e flow, runs preflight, submits the OSMO workflow, and retrieves the augmented dataset.

Quick Start

Ask your agent to run the full Image Attribute Augmentation pipeline on your person-crop dataset at its storage URL, specifying the OSMO pool and GPU platform.

Frequently Asked Questions about physical-ai-image-attribute-augmentation

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

FAQPage Schema
How do I run image attribute augmentation on NVIDIA OSMO?▼

Provide a person-crop dataset URL, then the skill selects a flow (e2e, augmentation, or auto_labeling), runs scripts/preflight_credentials.sh, verifies NIM endpoints, and submits the workflow with osmo workflow submit. Outputs are retrieved with osmo data download.

What dataset format does image attribute augmentation require?▼

The pipeline expects person-crop images organized as <person_id>/<view>.jpg subdirectories uploaded to OSMO storage. There is no built-in demo dataset, so you must supply your own dataset location before any run.

Which model endpoints does the augmentation pipeline need?▼

Three OpenAI-compatible endpoints are required: Qwen-Image-Edit-2511 for image editing, a VLM (qwen3-vl) for MCQ verification and captioning, and an LLM (qwen25-14b) for question generation. In-cluster NIMs are the default; external URLs are opt-in overrides.

Why does the image-edit endpoint return 404 on chat completions?▼

The Triton-based NIM image serves only /v1/infer and /v1/images/edits, not /v1/chat/completions. Deploy the vllm/vllm-omni image with vllm serve Qwen/Qwen-Image-Edit-2511 --omni, which exposes the required chat completions image-edit path.

Can I run only captioning without image augmentation?▼

Yes, select the auto_labeling flow, which runs setup plus the captioning worker on pre-augmented images. It uses only the VLM and LLM endpoints and produces open_qa.json captions from the person_attributes question bank.

Related Skills

Explore More Software Engineering Skills→
patriotnewsactivismpatriotnewsactivism
1

object-storage

Automate Google Cloud Storage bucket setup, file uploads, and access control.

Community
Intermediate
NVIDIA-TAONVIDIA-TAO
83

tao-train-reid

Train, evaluate, export, and run inference for TAO person re-identification models.

Official
Advanced
rish2jainrish2jain

nvidia-nim

Deploy and operate NVIDIA Inference Microservices for GPU-accelerated AI inference.

Community
Advanced
eachlabseachlabs
28

video-noise-reduction

Reduce noise and grain from video files using the each::sense AI API.

Official
Intermediate