VisionForge for dataset labeling

Local bounding-box labeling,
on your PC.

  • Create or open a .VFSln — one project file holds the name, image folder, labels, and how you annotate.
  • Draw and review boxes with Cursor and Box, then step through the folder with the playback bar.
  • Export YOLO or Pascal VOC next to your images, with a class list your training pipeline already reads.

Screens

Every screen, explained

Three views from the desktop app. Each preview follows the real VisionForge window — not a cloud dashboard.

UI preview

Frame: 3

Workspace

  • Left tools: Cursor, Box, Hexagon, Select Images, Process Image
  • Hexagon is on the rail only — it does not draw polygons
  • View tools and a playback bar under the image
  • Right inspector: Assets, Labels, Detections

How it works

Your five-step labeling path

From the start page to exported label files — a local path on Windows, not a cloud dataset tool.

  1. Start

    Create new project, Open existing project, or pick a recent .VFSln from the list.

  2. Create

    Set project name, folder, annotation type, and mode. Next writes the .VFSln.

  3. Open folder

    File → Select Image Folder fills Assets. The playback bar steps through the set.

  4. Label

    Add classes in Labels. Press W for Box or Cursor, then draw and review boxes.

  5. Export

    File → Export writes YOLO or Pascal VOC next to the images, plus a class list.

Features

What you can do in the desktop app

Built for a labeling desk on Windows. You draw bounding boxes today. Preview-only tools are marked clearly.

Image folder and playback

File → Select Image Folder loads common image types. The Assets list sets the current picture. Skip, step, play, and seek from the bar at the bottom.

Inspector: Assets, Labels, Detections

The right panel opens with the project. Import a classes.txt when labels are empty. Add, rename, and delete stay in the project file. Detections lists boxes on the current image.

Box draw and review

Press W to switch Cursor and Box. Click and drag to draw with the selected label. Cursor moves and resizes. Delete or Backspace removes a box.

Zoom, pan, rotate

Zoom in, Zoom out, Fit to Screen, and Rotate 90° clockwise. Rotate saves the turned image. Middle-button pan. Ctrl and the scroll wheel zoom.

Export and import label files

Export YOLO or Pascal VOC plus a class list. Existing label files import only for images that do not already have boxes. Reviewed work stays.

Process Image Preview only

Pick a model file and a class list from the left tools. Process Image does not run the model yet.

Inspector

Assets, Labels, and Detections in one panel

The right inspector opens after a project is open. Start on Assets, then manage classes and review boxes on the current image.

UI preview

Assets Labels Detections
  • img_001.jpg
  • img_002.jpg
  • img_003.jpg
  • img_004.jpg
Labels Add
  • person
  • car

Labels stay in the project

If the project has no labels yet, VisionForge can import a classes.txt from the project folder or the image folder. Add, rename, and delete save in the .VFSln. Each class gets its own color.

Export

What you receive next to the images

File → Export writes label files your training pipeline already reads. The project type stays as you created it. You choose YOLO or Pascal VOC for that run.

UI preview — sample export

Label files beside the image folder

Destination starts as your image folder · type locked · mode choosable for this export

YOLO

One text file per image, plus a class list. Ready for a YOLO training folder.

Pascal VOC

One XML file per image, plus a class list. Ready for a VOC-style dataset.

Import on open

Existing label files load only for images that do not already have boxes. Work you already reviewed is not overwritten. New images still join the list.

Softasium-licensed Windows desktop

Get started in three steps

VisionForge is a local annotator for teams who already use YOLO and Pascal VOC and want the project in one .VFSln file.

  1. Download the installer

    Download VisionForge Release LTS for Windows and run the installer.

  2. Install on Windows

    Install the desktop app on the machine that holds your images.

  3. Create a project

    Create new project, choose a folder, then Select Image Folder and start drawing boxes.

Disclaimer: Drawing today is bounding boxes only. Process Image is a preview screen and does not run a model. Hexagon appears on the tools rail; it does not draw polygons.

Requires Windows. Windows may show a SmartScreen prompt for unsigned installers. Licensed through Softasium (www.softasium.com).