Vis-O-Matic is a free and easy to use suite of tools for analyzing image datasets: use Image Cutter to segment parts of images using computer vision; use Batch Image Analyzer to measure the colour properties of image pixels and clusters; and use Image Visualizer to and explore and interpret image colour datasets.
View, arrange, and organize image datasets
Isolates part of an image using a transformer model (Segment Anything from Meta AI) that runs locally in the browser. Click what you want, drag a box around it, or let the model propose every region it can find.
Reads colour across a whole set of images at once. Clusters each image into its dominant colours and reports values in HSV, LAB and LCH alongside whole-image summary metrics.
Turns colour values into something you can read at a glance. Paste colour data or load a CSV, then arrange them using different visualizations (strip, grid, scatter, wheel or three-dimensional plot).
A typical sequence runs left to right. Add your images, cut a region out of each image, and export the cutouts as transparent PNGs. Open that folder in the Analyzer so the colour figures describe only the region you kept rather than the whole frame. Take the resulting cluster CSV into the Visualizer to see how the palette is distributed.
All tools can be used independently. The Analyzer also reads ordinary photographs perfectly well without the Cutter, and the Visualizer accepts colour values from any source, including data that never passed through either of the other tools.
They run on your machine, in a browser tab, which sets the limits. Very large image sets may make the browser sluggish. The Cutter downloads a model of between 20 and 300 MB the first time it runs, and needs WebGPU to avoid longer processing times, where possible.
The segmentation model proposes a mask; it does not verify one. Models inherit the biases of their training data and perform unevenly outside the settings they were trained on. Treat a mask as an instrument reading, report which model produced it, and check a sample by eye. Every export records the model and the device it ran on.
No image ever leaves your computer. There is no server, no account and no upload step. The tools work offline. The Image Cutter only requires Web access once to download the Segment Anything model, which is then cached locally.
If these are useful in research or teaching, please cite the specific tool:
Each tool is a single HTML file released under the MIT licence. Model weights and libraries carry their own licences, noted in each tool.
Vis-O-Matic sits alongside a few other visual research tools: