Import data

Prepare a small import

Use data you own or are authorized to process. Remove secrets and unnecessary personal information. Keep an original copy outside Lutrio. For object detection, decide the class names before labeling.

Import into a CV project

1. Open the project’s dataset area and choose the add/upload action, when offered.
2. Select the images or archive accepted by the dialog. Read the displayed file-type and size limits; they can differ by service configuration.
3. Submit once and follow the upload/import task. Large files may continue processing after upload reaches 100%.
4. Wait for the final result, then compare imported and skipped counts with your source. Open several images to check orientation and quality.

The dataset action may be labeled 添加数据集 (add dataset). Existing-dataset association and new upload are different actions: association references data already available to that project.

Archives and incomplete uploads

Use the format accepted by the current upload dialog. If an archive imports fewer images than expected, inspect the final task result before uploading it again. Non-image files may be ignored. A successful transfer alone does not prove every sample was imported.

VLM data

VLM data needs images together with the corresponding questions, instructions or answers. Use that project’s preparation/publishing workflow rather than the CV upload dialog. Confirm the accepted schema with the team before uploading a large batch.

What to check next

Raw images are not yet a reproducible training version. Inspect labels, review the dataset and publish the version used for training. Continue to annotation and review.