Annotation Studio

Most annotation tools want you to upload your dataset to their cloud. That's fine for a school project. It's a non-starter when you're working with medical scans, security footage, customer photos, or any data you can't legally share with a third party.

Annotation Studio takes the opposite approach: nothing leaves your machine. You install it, point it at a folder of images on your disk, and start labeling. The app reads images from disk and writes YOLO-format .txt files right next to them — exactly the layout YOLOv8/v11 expects when you train.

Open-source alternatives like CVAT and LabelImg ask you to install Docker, set up Python environments, manage dependencies. Annotation Studio is a single Windows installer. Double-click to install. The app handles everything else — including a small embedded Python runtime that powers AI auto-annotation when you ask for it.

The interface is built for speed: number keys to switch classes, click-and-drag to resize boxes, keyboard shortcuts for everything, dark mode by default. Pose keypoints are first-class — draw the box, click each keypoint, hit Enter.

Have a YOLO .pt model? Drop it in. The app runs inference locally on your GPU (NVIDIA + CUDA) or CPU, writes the predictions into your .txt files, and lets you tweak them.

The first time you use this feature, the app downloads PyTorch + Ultralytics directly from PyPI to your local app data folder — about 350 MB for CPU, 2.7 GB for GPU. After that, you're fully offline.