Abstract

For Insect Classification there are many methods proposed. To find the more suitable classifier we have implemented two different methods of classification, Custom CNN and Transfer Learning. We observed the accuracy and loss parameters during training phase and validation phase on both Custom CNN and Transfer learning methods. In Transfer Learning we have created the base model from the pre-trained model MobileNetV2. This model is further trained on Imagenet Dataset which consists of 1.2M labeled images .We compared all attributes of both models on this dataset. Recorded Accuracy and Loss for training and Validation. From the results we found that Custom CNN model works better than Transfer Learning for this dataset. This model can also be used for education and farming purpose.