A sufficient training dataset for object detection includes:
- Image files (same jpg, png, or bmp format); recommended 640 x 640 -pixel image at minimum
- Label files for each image (TXT format)
- Map file (ANS_Class.json) that contains label IDs for the software to identify
- The training dataset must be placed inside a folder named "data" to meet the ANSTS condition.
ANSTS also includes a built-in ANS Labelling Tool:
- If your dataset doesn't have label and map files, you can use the Labelling Tool to create your own label and map files.
- If your dataset already has label and map files in different formats (XML, JSON, TXT), the Labelling Tool supports importing those files so you can quickly reuse your label information.