Function Description
The tag defect detection includes the following steps: detection, segmentation, edge defect filtering, workpiece alignment, and center defect classification. The detection and segmentation datasets in COCO format can be directly labeled on an AI platform, but images in the classification dataset must be workpiece-aligned. The process of creating a dataset for tag detection and classification is as follows:
- Split a dataset. Retain only the tag images without defects in the images folder and perform the operations described Tag Extraction Tool in to obtain images whose classification results are ok.
- Perform operations described in Data Augmentation Tool for Tag Wrinkles on the ok images to obtain images with wrinkles and bubbles.
- Upload the three types of images to the AI platform and create three datasets corresponding to the ok, wrinkle, and bubble images.