Upload to Train
Upload To Train

1 · Add your data ?

Upload a file or load a sample. Supported: .csv / .tsv / .txt (tabular & text) and .zip of class-foldered images.

⬆️

Drop files here, or click to browse

several at once · max 200 MB each

🖼️ How to prepare an image dataset show

What's inside an image dataset. It's a single .zip file. Each top-level folder = one class label, and the image files inside that folder are the examples for that class. The folder name becomes the label the model learns to predict.

Required ZIP structure

my_images.zip
├── cat/            ← class "cat"
│   ├── 001.png
│   ├── 002.jpg
│   └── 003.png
├── dog/            ← class "dog"
│   ├── 001.png
│   └── 002.png
└── rabbit/         ← class "rabbit"
    ├── 001.jpg
    └── 002.png

How to create it (3 steps)

  1. Make one folder per category and name it exactly as the label.
  2. Drop that category's images into its folder (≥ 10 per class; more is better).
  3. Select all the folders → right-click → Compress / Send to → ZIP, then upload the .zip and set Treat as: image.
  • Accepted image types: .png, .jpg/.jpeg, .bmp, .webp. Any size — they're auto-resized.
  • Use 2+ classes, kept reasonably balanced (similar image counts per class).
  • Don't add a parent wrapper folder — the class folders must be at the top level of the zip.
  • No dataset yet? Pick any 🖼️ Image sample on the right (incl. the new domain sets) to see the exact format in action.

Or load a sample dataset

Your datasets ?