Cara.app Art Dataset — 123,056 Original Images from Top Community Posts (120 GB)
Yuskeu

@article{,
title= {Cara.app Art Dataset — 123,056 Original Images from Top Community Posts (120 GB)},
journal= {},
author= {Yuskeu},
year= {},
url= {},
abstract= {# **companion metadata database**: https://academictorrents.com/details/cab03c1b0cbc8ed99a0495d491d281d80517146a

This archive contains 123,056 original artwork image files (~120 GB, JPG/PNG/GIF/WEBP) belonging to the 52,047 most-liked posts (over 150 likes each) on Cara.app, an art-sharing and portfolio platform popular among professional and hobbyist illustrators. Images were collected in August 2026 from Cara's public content-delivery network and are provided at original uploaded resolution. Each file is uniquely named by its CDN path and maps one-to-one to rows in the companion metadata database (`subset_dataset.db`, distributed separately) via its filename. 

The **companion database**: https://academictorrents.com/details/cab03c1b0cbc8ed99a0495d491d281d80517146a (Cara.app Art Dataset — Full Metadata Database for 52,045 Top Posts with 674k Comment Trees (SQLite)) provides per-image carousel ordering, cover flags, pixel dimensions where recoverable, platform AI-generation flags, full post text and engagement metrics, complete comment threads, and author profiles. Intended uses include computer-vision research, art-market and style-trend analysis, AI-vs-human classification studies, and digital-arts historiography.

---

## How to link image files to metadata (filename lookup)

Every file in the `images` folder maps to exactly one row in the `images` table by its **basename**, which carries `post_id` — from there you reach the full post and author.

**SQL (any sqlite3 client):**
```sql
-- all metadata for one downloaded file:
SELECT i.*, p.title, p.content, p.created_at, p.like_counter,
       u.name AS author_name, u.slug AS author_slug, u.follower_counter
FROM images i
JOIN posts p ON p.id = i.post_id
JOIN users u ON u.id = p.author_id
WHERE i.cdn_url LIKE '%/' || 'tunamelt-xVmO9g6D26ViDKkmTZnOz-7d865f66-b294-4e12-a0fd-8099137f8467.jpg';

-- build the complete file→post→author manifest yourself:
SELECT substr(i.cdn_url, length('https://cdn.cara.app/')+1) AS filename,
       i.post_id, p.created_at, p.like_counter, u.slug AS author_slug,
       u.name AS author_name, i.is_cover, i.ord, i.width, i.height
FROM images i JOIN posts p ON p.id=i.post_id JOIN users u ON u.id=p.author_id;
```

**Python (3 lines):**
```python
import sqlite3
db = sqlite3.connect("subset_dataset.db")
fn = "tunamelt-xVmO9g6D26ViDKkmTZnOz-7d865f66-b294-4e12-a0fd-8099137f8467.jpg"
img  = db.execute("SELECT * FROM images WHERE cdn_url LIKE '%'||?", ("%/"+fn,)).fetchone()
post = db.execute("SELECT * FROM posts WHERE id=?", (img[1],)).fetchone()
```
(`images.post_id` is column 2; `posts.author_id` is column 2; `users.id` is column 1.)

The same join works for comments: `comments.post_id → posts.id`.},
keywords= {machine learning, Computer Vision, cara.app, art dataset, digital art, illustration, image dataset, generative AI detection, artists, social platform},
terms= {},
license= {},
superseded= {}
}