Classification of Corn Seed Quality using Residual Network with Transfer Learning Weight

Corn is one of the main ingredients in farm animal feed. Currently, corn preferable because widely available and cheaper market than others. However, it needs quality control on production. The company that manufactures feed has certain standards to receive material. On other hand, produced varies greatly. Thus, when receiving from suppliers greatly affects classified into physical properties analytical values. Physical are determined so resulting can be accepted or rejected, while value used as basis for formulating diet. by human senses, such sight smell, chemical analysis. relying senses certainly limited takes time. Based these problems, make a classification system seeds automatically. This study uses seed images data. public data Naagar which consists four classes: pure, discolored, silk cut, broken. Image Convolutional Neural network (CNN) with ResNet152v2 architecture. hyperparameters consist learning rate 0.001, batch size 512, an epoch 25. Adaptive Moment Estimation (Adam) optimizer. Percentage training vs validation 80:20. results show accuracy 65%, precision 66%, recall 64%.