This pulp modeling inspection line includes feeding, inspection, counting, and stacking together to solve the heavy labor cost problem in the current industrial. The system can detect both inside and outside defects including but not limited to iron wire, dirt, light transmission, burrs, crushing, damage, etc. with 10,000pcs/h throughput. The inspection section is powered by deep learning AI algorithm, with strong software and hardware compatibility, and can detect most common product, various from boxes, bowls to burger boxes and other products. Defects can be accurately classified and counted, so as to monitor the previous process. The system has the ability to learn and getting stronger during usage, and both OK and NG product can be stacked.