<![CDATA[Artificial intelligence was the powerful approach that was proven to be impactful for solving several problems. In the leather inspection cases, artificial intelligence also contributed some research works that effected for leather inspection process. In this research, we employed NasNet architecture conducted by using fine-tunning transfer learning method to distinguish the types of leather defects. We used 3600 images that was distributed into six classes which are folding marks, grain off, growth marks, loose grains, pinhole and non-defective. Our proposed solution successfully achieved accuracy for training data is 0.9788 with loss of 0.0198. While the maximum accuracy in validation data is 0.8059 with loss of 0.2126. In the testing data, our experiment obtained accuracy of 0.8603 with loss of 0.1603. These results indicated that our proposed solution was suitable to recognize the characteristics of leather defects and suitable to distinguish them.]]>
Copyrights © 2023
