Use Your Own Dataset and Model
Prepare compatible inputs and import them as Project-owned Materials.
Knotest evaluates a fixed Dataset with an Inference Source. For a local Model, the Dataset task, Model task, output classes, and inference settings must agree.
Choose the task first
Do not begin by choosing a file extension. First identify the scientific task, then prepare a Dataset format and Model configuration that implement that task.
Classification
Image-folder classification inputs.
Object Detection
YOLO or COCO bounding-box inputs.
Segmentation
YOLO polygons or COCO instance segmentation.
Oriented Bounding Boxes
YOLO four-corner oriented boxes.
Tabular
CSV, TSV, or Parquet classification and regression.
Import creates Project ownership
Import copies the selected source into the Project. Knotest records its content identity and does not continue using the original external path. Keep the original source as your own backup, but treat the imported Material as the Evaluation input.
Before importing, review the complete supported format table.