Workflow
Projects & data
A project is a self-contained workspace: your media, labels, annotations, and training runs all live together — on disk, on your machine.
When you open VisoLabel you land on the projects screen. Recent projects show as thumbnail cards with their image count and task type, so you can pick up where you left off or start something new.
Creating a project
Click New project, give it a name and optional description, then pick a task type. VisoLabel creates a project folder on disk that holds a small database of your annotations plus references to your media — your original files are never moved or modified.
Task types
The task type sets sensible defaults for annotation and training:
- Object Detection — locate objects with bounding boxes. Good for counting and tracking (Bounding Boxes, Counts, Tracking).
- Instance Segmentation — capture each object's actual shape with pixel-level polygons (Polygons, Measuring, Odd Shapes).
- Classification — assign a label to the whole image, with no boxes or polygons (Image Labels, Filtering, Content Moderation).
Importing media
You can add data in a few ways:
- Drag and drop a folder or files onto the window.
- Import → Add files to pick individual items.
- Import → Add folder to bring in a whole directory, recursively.
Supported formats
| Type | Formats |
|---|---|
| Images | .jpg, .jpeg, .png, .bmp, .webp, .tiff |
| Video | .mp4, .mov, .avi, .mkv |
Videos are split into frames on import so you can annotate frame by frame. Object tracking carries labels across frames — see the Annotation guide.
Organizing your dataset
The media panel lets you filter and sort to find what needs attention:
- By status — unlabeled, in progress, or reviewed.
- By label — show only items containing a given class.
- Flagged — mark tricky items to revisit later.
Importing existing annotations
Already have labels from another tool? Import them so you don't start over. VisoLabel reads common formats:
- COCO JSON
- YOLO txt + data YAML
- Pascal VOC XML
Where your data lives
Everything stays local. The project database and any exported datasets are written to the project folder you chose at creation. Nothing is sent to a server — that's the point of local-first. Back up the project folder like any other directory.
When your images are labeled, you can review them all in the gallery and expand your data with augmentation — see Datasets & augmentation.