MuhammadLab
Computer VisionBrowser-basedDataset creationBoxes and masksYOLO / COCO / CSVStudent lab

Image Annotation Tool

Upload an image, draw bounding boxes, polygons, or brush masks, assign class labels, and export annotations in common training-data formats.

This studio teaches how computer vision datasets are created before model training. Students can compare coarse object boxes with more precise polygon and mask annotations, then inspect how those choices map into YOLO, COCO, and CSV files.

Why annotation matters

Before a model can learn to detect or segment anything, humans usually create the training labels first. Boxes, polygons, and masks become the ground truth a model is asked to imitate.

What students practise

This page lets students create bounding boxes for detection, polygons for segmentation boundaries, and brush masks for pixel-level regions, then export the labels in common dataset formats.

What to compare

Boxes are quick but coarse, polygons trace shape boundaries, and masks capture pixel-level detail. Seeing all three makes the tradeoff between speed and precision much easier to understand.

Upload an image to start drawing training annotations.

Boxes

0

Polygons

0

Masks

0

Exports

Download the dataset labels

YOLO exports one normalized box per line, COCO exports a structured JSON file with categories and annotations, and CSV gives a simple table that students can read directly.

Class Labels

18px

Annotation List

No annotations yet. Draw a box, polygon, or mask after uploading an image.

Learning Notes

  • Bounding boxes are common for object detection because they are fast to create and easy to train on.
  • Polygons trace shape boundaries more precisely, which is helpful when students need region-level segmentation labels.
  • Brush masks are closest to pixel-wise segmentation, but they take the most annotation time.
  • Exporting the same scene in YOLO, COCO, and CSV helps students see how one visual annotation becomes structured training data.