Based on the large number of banana production carried out by small and medium entrepreneurs, their processing is still carried out inefficiently. The inefficient process can be seen from the selection of the types of bananas to be processed, which are still done manually, by looking at the characteristics of color, shape and texture with the naked eye. Thus, an alternative is needed to reduce or overcome problems for its selection, sorting and utilization. For this reason, a technology is needed that can support the process of selecting, sorting and identifying various kinds of bananas, so that they can help small and medium entrepreneurs, including manufacturers, in processing and selecting various kinds of bananas. This technology will make it easier for users to detect and recognize the variety of bananas, including providing information about the benefits provided to each type of banana. The technology used is artificial intelligent technology using the concept of Machine Learning to support deep learning algorithms. So that in this research several convolutional neural network models will be developed which are derivatives of deep learning algorithms to be able to classify and detect various types of bananas automatically. To improve classification accuracy and optimize detection, this research consists of several stages starting from the initial process, feature extraction and classification or image detection of bananas. In the initial stages, the Gaussian Blur method will be used, edge detection and sharpening of the image using SUCK (Sharpening using Custom Kernel) or Contrast Limited adaptive histogram equalization (CLAHE). This initial process was carried out to support the process of feature extraction and classification by using several Convolutional Neural Network (CNN) models and the support of several optimizer methods. Meanwhile, the image detection process will also besupported by the darknet algorithm. In collecting data other than various banana images and several other fruit images which have almost the same image characteristics, it is hoped that a reliable and robust system will be formed against object changes. All methods or algorithms are used in order to make it easier for users to recognize and detect several types of various banana fruit images optimally.

Banana, CNN, SUCK, CLAHE, Darknet Algorithm, optimer

• Ranny, S.Kom., M.Kom.

• Dr. Abdul Haris Rangkuti, S.Kom., M.M., M.Si.

• Dr. Rudy Aryanto, S.E., M.M.