Update to Roboflow

{
  "description": "Use when searching for or using public datasets/models on Roboflow Universe (universe.roboflow.com), the open repository of 1M+ computer vision datasets and 50K+ pre-trained models.",
  "included_files": [],
  "name": "roboflow-universe",
  "skill_md_contents": "---\nname: roboflow-universe\ndescription: Use when searching for or using public datasets/models on Roboflow Universe (universe.roboflow.com), the open repository of 1M+ computer vision datasets and 50K+ pre-trained models.\n---\n\n> **For agents:** This skill is authored in [`roboflow/computer-vision-skills`](https://github.com/roboflow/computer-vision-skills). If your client already has it loaded as a skill (for example from a Roboflow plugin), use that copy. The `roboflow://skills/<name>/...` MCP resources serve the same skills for clients that don't have them loaded.\n\n# Roboflow Universe\n\nOpen repository of 1M+ computer vision datasets and 50K+ pre-trained models at `universe.roboflow.com`.\n\n## URL Patterns\n\n| Page | URL | Content |\n|------|-----|---------|\n| Home | `universe.roboflow.com` | Search, trending projects, categories |\n| Project | `universe.roboflow.com/{owner}/{project}` | Overview, classes, metrics, license, fork |\n| Images | `universe.roboflow.com/{owner}/{project}/browse` | Browse images with annotations |\n| Dataset version | `universe.roboflow.com/{owner}/{project}/dataset/{version}` | Version details, splits, download |\n| Model | `universe.roboflow.com/{owner}/{project}/model/{version}` | Try model, metrics, deploy snippet |\n\n## Searching Universe\n\n### MCP Tool\n\nUse `universe_search` to find datasets/models programmatically. Pass a descriptive query (e.g. \"hard hat detection construction site\").\n\nWhen someone must **choose a dataset after seeing it** (previews, classes, license, image counts), share the Universe links from the results so they can browse the listings before deciding.\n\n### Web Search\n\nSearch is **hybrid** — combines semantic similarity with keyword matching. Use specific, descriptive queries for best results.\n\n### Query Operators\n\nAll operators can be mixed with free-text: `fire smoke class:fire,smoke images>200 model`\n\n| Operator | Example | Effect |\n|----------|---------|--------|\n| `model` | `waste detection model` | Only datasets with a trained model |\n| `object detection` | `helmet object detection` | Filter by project type (also: `classification`, `instance segmentation`, `keypoint detection`) |\n| `class:X` | `class:helmet,person` | Must contain these classes |\n| `tag:X` | `tag:safety` | Filter by Universe tag |\n| `model:X` | `model:yolov8` | Filter by trained model architecture |\n| `images>N` | `images>500` | Min image count (also `>=`, `<`, `<=`, `=`) |\n| `stars>=N` | `stars>=5` | Min star count |\n| `views>N` | `views>1000` | Min view count |\n| `downloads>N` | `downloads>100` | Min download count |\n| `updated:Nd` | `updated:30d` | Updated within N days (also `h`, `w`, `mo`, `y`) |\n| `sort:X` | `sort:stars` | Sort by field (stars, images, updated, downloads, views) |\n| `like:dataset-url` | `like:coco` | Find similar datasets |\n\n### Tips for Effective Queries\n\n- Combine free-text with operators: `pothole road damage class:pothole images>100 sort:stars`\n- Add `model` to only get inference-ready datasets\n- Include project type keywords to filter: `helmet instance segmentation`\n- Use `class:` when you know exactly what classes you need\n- Use specific object names, not generic terms (\"forklift in warehouse\" > \"vehicle\")\n\n## Evaluating a Dataset\n\nBefore forking, check these signals:\n\n| Criterion | Where to Look | Good Sign |\n|-----------|---------------|-----------|\n| Class coverage | Classes list on project page | All your target classes present |\n| Image count | Project overview | 500+ images per class for detection |\n| Annotation quality | Browse > click individual images | Tight bounding boxes, consistent labels |\n| Class balance | Project overview / health | No extreme class imbalance |\n| Image diversity | Browse images | Varied lighting, angles, backgrounds |\n| License | \"Cite this Project\" section | Compatible with your use case (see below) |\n| Model metrics | Model tab (if available) | mAP > 70% suggests decent annotations |\n\n## Licenses\n\nFound in the \"Cite this Project\" section on the project page. No license listed = all rights reserved.\n\n| License | Commercial Use | Modify | Attribution Required |\n|---------|---------------|--------|---------------------|\n| Public Domain | Yes | Yes | No |\n| CC BY 4.0 | Yes | Yes | Yes |\n| MIT | Yes | Yes | Yes (in license copy) |\n| BY-NC-SA 4.0 | No | Yes (share-alike) | Yes |\n| ODbL v1.0 | Yes | Yes (share-alike for DB) | Yes |\n| No license specified | Assume No | Assume No | N/A |\n\n## Forking a Dataset\n\nFork = copy a Universe dataset into your workspace (no download/re-upload needed).\n\n1. Open dataset on Universe\n2. Click **Download Dataset** button\n3. Choose **Train a model with this dataset** (full fork) or **Train from a portion of this dataset** (partial clone)\n4. Dataset copies into your workspace\n\nAfter forking you can: rename classes, add/remove images, generate versions, train models.\n\n**Requires:** Logged-in Roboflow account.\n\n## Downloading a Dataset\n\nFor local/notebook training instead of Roboflow cloud training.\n\n| Method | When to Use |\n|--------|-------------|\n| Train a model with this dataset (fork) | Training on Roboflow, want full data in workspace |\n| Train from a portion (clone) | Want a subset or to combine with other data |\n| Download dataset | Local training via code snippet or ZIP file |\n\nSupports all standard export formats (COCO, YOLO, VOC, CreateML, TFRecord, etc.).\n\nPath: Project page > **Download Dataset** button > choose method.\n\n## Using a Universe Model\n\n### Direct Inference via Workflows\n\n1. Create a Workflow in Roboflow\n2. Add a model block\n3. Switch to **Public Models** tab\n4. Paste the model ID from the Universe model page (copy icon at top)\n5. Click **Use model ID**\n\nModel ID format: `{owner}/{project}/{version}` (shown on Universe model page).\n\n### Checkpoint Training\n\nFork the dataset, then train your own model using the forked data. Use a Universe model's architecture as a starting point via Roboflow Train.\n\n### Inference Metrics (shown on model page)\n\n| Project Type | Metrics Shown |\n|--------------|---------------|\n| Object Detection | mAP, precision, recall |\n| Classification | Accuracy |\n| Segmentation | mAP, precision, recall |\n\n## MCP Tool Reference\n\n`universe_search` — Search Universe for datasets/models.\n\n| Param | Type | Default | Notes |\n|-------|------|---------|-------|\n| `query` | str (required) | — | Search query text |\n| `result_type` | `\"dataset\"` \\| `\"model\"` \\| null | null | Filter by result type |\n| `limit` | int | 12 | Max results per page |\n| `page` | int | 1 | Page number (1-indexed) |\n\nReturns: name, url, type, classes, classCount, images, description, tags, license, stars, views, downloads, modelCount, latestVersion.\n\n## Related Skills\n\n- `roboflow://skills/roboflow-data-management/SKILL` — managing datasets after import\n- `roboflow://skills/roboflow-training-and-evaluation/SKILL` — training on forked data\n"
}