Shiro(ሽሮ), is commonly used traditional food in Ethiopia. It is a typical Ethiopian ethnic staple meal that is eaten throughout the country. Shiro flour variety can be prepared from different types of legumes according to societies living style. Some commonly used types of legumes for preparing Shiro flour variety are vetch (ጏያ), chickpeas (ሽንብራ), peas (አተር), and it can be prepared by mixing two or more legumes in one we call as it a Mixed (ድብልቅ). Most Shiro flour variety products based on the particular Baltena Shop marketplace have similar Shiro flour characteristics (colors, ingredients, and texture). Because of this reason, it is difficult to identify them using the human eye which one is pure and the mixed one. Even if it is identified by the human eye the identification process is very tedious and time-consuming. Although there are no research works that support the manual Shiro flour variety classification process automatically. So, the objective of this study is to develop an automatic Shiro flour variety recognition model using a combined features of imaging and machine learning techniques. A total of 6000 Shiro flour images were collected from different particular Baltena Shop marketplace by using a mobile camera. After getting the dataset, image preprocessing tasks like Histogram equalization, Noise reduction are and other are done. In feature extraction, we apply, LBP, GLCM, CNN, and the combination of CNN (automatic feature extraction) and GLCM (traditional feature extraction). It is used to select the important features that account for the classification of the Shiro flour variety. For classification, we have used Convolutional Neural Network (CNN), Support Vector Machine (SVM), and, K-Nearest Neighbors (KNN). And finally, we used python as a programming language with Keras library (TensorFlow backend). Our CNN model achieved 82% testing accuracy. SVM model using each of the feature extractors that we have used as ( i.e., CNN, LBP, GLCM) feature vectors achieved, 98.8%, 68.8%, 77.3%, accuracy respectively and with the combined feature vectors of CNN and GLCM, it was achieved 99.7% of accuracy. KNN model using each of the feature extractors, (i.e., CNN, LBP, GLCM) feature vectors achieved 98.9%, 76.8%, and 93.5%, accuracy respectively. Finally, the combined feature vectors of CNN and GLCM with the KNN classifier was achieved 99.9% accuracy. This combined feature extractor outperforms the best accuracy from the others for classifying Shiro flour varieties. Keywords: Digital Image Processing, Image preprocessing, Feature extraction techniques, Ethiopian Shiro production.