Dataset Card
COCO-CLS-5k
A public computer-vision dataset of 5,000 images with 36,335 annotations across 135 classes, annotated as polygon, bounding box, and tag.
Pictograph Research · a custom license · Public · Updated 2026-11-03
Overview
This dataset holds 5,000 images annotated as polygon, bounding box, and tag across 135 classes. The most common auto-detected tags are playing, eating, person, eye-level view, top-down view, zoo. It is ready to import into a workspace, export in a dozen formats, or train a model on.
At a glance
| Metric | Value |
|---|---|
| Images | 5,000 |
| Annotations | 36,335 |
| Classes | 135 |
| Annotation types | polygon, bounding box, tag |
| Folders | 1 |
| Storage | Standard |
| Visibility | Public |
What's in these images
Tags auto-detected across the dataset, most common first.
| Tag | Category | Images |
|---|---|---|
| playing | Attributes | 696 |
| eating | Attributes | 563 |
| person | Objects | 471 |
| eye-level view | Attributes | 414 |
| top-down view | Attributes | 340 |
| zoo | Scenes | 248 |
| sign | Objects | 235 |
| walking | Attributes | 219 |
| driving | Attributes | 212 |
| child | Objects | 211 |
| bus | Objects | 203 |
| cooking | Attributes | 197 |
| jumping | Attributes | 197 |
| bathroom | Scenes | 195 |
| sitting | Attributes | 194 |
Image tags
Every image carries exactly one image-level tag: 55 tags across all 5,000 images, so the dataset trains an image classifier as it is. The tags were assigned from the dataset's SigLIP embedding map and reviewed by eye; other holds the 313 images no other tag fits. The COCO polygons are kept alongside.
| Tag | Images |
|---|---|
| street | 228 |
| meal | 176 |
| bathroom | 175 |
| cat | 167 |
| tennis | 164 |
| skiing | 160 |
| kitchen | 149 |
| train | 143 |
| baseball | 137 |
| bus | 137 |
| skateboarding | 133 |
| livestock | 131 |
| pizza | 130 |
| surfing | 128 |
| clock | 127 |
| dining | 126 |
| office | 110 |
| horse | 108 |
| living room | 106 |
| motorcycle | 104 |
| giraffe | 102 |
| dog | 95 |
| airplane | 90 |
| video games | 89 |
| elephant | 84 |
| formal wear | 84 |
| kite | 81 |
| fruit | 78 |
| portrait | 76 |
| zebra | 76 |
| street sign | 74 |
| boat | 71 |
| teddy bear | 71 |
| frisbee | 65 |
| fire hydrant | 57 |
| bird | 53 |
| bedroom | 52 |
| bear | 48 |
| vase | 43 |
| bench | 42 |
| cake | 40 |
| child | 40 |
| donut | 40 |
| luggage | 39 |
| electronics | 38 |
| market | 33 |
| stop sign | 31 |
| vegetables | 29 |
| hot dog | 27 |
| dessert | 24 |
| beach | 23 |
| kitesurfing | 22 |
| soccer | 18 |
| vintage photo | 13 |
| other | 313 |
Use this dataset
from pictograph import Client, Dataset
client = Client() # reads PICTOGRAPH_API_KEY
dataset: Dataset = client.datasets.get("COCO-CLS-5k")
print(dataset.image_count, dataset.annotation_count)- Import it into your own workspace with one click from its page.
- Pull it with the Pictograph SDK (shown above) or the CLI.
- Export to COCO, YOLO, Darwin V7, Pascal VOC, and more.
- Train a model directly from an export.
License
Released under a custom license.
Source
The images and annotations come from COCO 2017, created by COCO Consortium (Lin et al., ECCV 2014) and released under CC-BY-4.0. Please credit the original authors when you use this data.
Source: COCO 2017, created by COCO Consortium (Lin et al., ECCV 2014). Credit the original authors when you use this data.
Non-commercial use only. Annotations: CC BY 4.0, the COCO Consortium. Images: the COCO Consortium does not own their copyright; each image comes from Flickr under its own license, recorded per image in COCO's annotation files, and most of them (68.9% of train2017) are Creative Commons NonCommercial. Use of the images must abide by the Flickr Terms of Use and each image's license. Cite Lin et al., Microsoft COCO: Common Objects in Context (ECCV 2014).
Models trained on this dataset
More from Pictograph Research
Build your first vision tool
$5 of free compute every month. No card required.