MLToolslabel · build · tune · ship

Annotation

Image Classification Labeler

For classification datasets speed is everything. Load a folder, define classes, and press a number key per image; the next image appears immediately. Export a CSV of file name and label, or a zip with one folder per class, the layout torchvision, Keras and timm read without code.

Loading the tool…

How it works

01

Load and define classes

Drop images and type the class names. Up to 9 classes map to number keys; more are picked from the list.

02

Label fast

One key per image. Space skips, Backspace goes back, and the progress bar shows what is left. Labels save automatically.

03

Export

labels.csv (filename,label) for pandas, or dataset.zip with class folders for ImageFolder-style loaders.

Formats, limits and the numbers that matter

Multi-label

Switch the mode to multi-label to allow several classes per image; the CSV then holds a semicolon-separated list.

Review

Filter by class to review every image with that label in a grid; click any to relabel.

Large sets

Thumbnails are generated at 256 px for the grid; the export copies the original files unchanged.

Common questions

Can I start from an existing CSV?

Drop the CSV alongside the images and matching file names are pre-labelled.

Does it resize images?

No. Use Batch Resize & Augment afterwards if the training pipeline wants a fixed size.

How do I split into train and validation?

Export the zip and open it in the Dataset Splitter, which keeps class folders and stratifies.

Next in the workflow

Train / Validation / Test SplitterSplit images with labels, class folders, or CSV/JSONL rows into train, val and test with a fixed seed and stratification.Batch Resize & AugmentResize, letterbox, flip, rotate and adjust a folder of images in the browser, with YOLO labels transformed to match.Duplicate Image FinderFind exact and near-duplicate images with perceptual hashing and remove them from a dataset.