Prepare Training Dataset

A sufficient ANSTS training dataset for object detection includes:

• A main dataset folder: A folder named data will contain all your dataset
• Image files (same jpg, png, or bmp format); recommended 640 x 640 -pixel image at minimum
• Label files: contain object classes information of each image (TXT format)
• Map file (ANS_Class.json): contains object class IDs for the software to identify

If your dataset is insufficient

ANSTS also includes a built-in ANS Labelling Tool:

• If your dataset only contains images, you can use the ANS Labelling Tool to start labelling.
• If your dataset already has label and map files but in different formats (YOLO, PASCAL VOC, COCO), the ANS Labelling Tool supports label conversion so you can quickly reuse your label information.

Example

Example dataset can be downloaded at Roboflow