Testing requirements:
(1) Detect surface defects on the inner and outer arcs of the magnetic tile, including cracks, chipping, grinding, and impurities.
(2) Detect 2 beats per second

Summary: Due to the influence of inherent pattern differences on the surface of magnetic tiles, crack detection requires high requirements for deep learning algorithms. Therefore, the products are classified into three categories: coarse crack sample A, fine crack sample B, and coarse fine good product C, in order to establish and optimize the model, which has a better detection effect on good products.