Adulteration of injera is produced by the addition of abuses, suggesting that the quantity of injera is increased by the addition of crops at low cost or things that can be added to get a high profit or decrease the cost, such as corn, Dagussa, gypsum, sawdust, mold injera, and others. In Ethiopia, most population of all ages uses injera as a basic food, and it is one of the most important sources of neutrinos. However, it has been noticed that some people purposefully add certain compounds or crops to pure teff. Adulterants are bad for human health, and it is difficult to identify them with human sensors. So, the detection of injera adulteration is an important factor in the success of a food product, particularly as food businesses aim to remain competitive in the global marketplace. Previous researchers addressed the problem through imaging techniques with traditional and state-of-the-art machine learning algorithms. Despite various attempts, the problem remains fully unsolved due to poor model performance and bad data collection methods, preprocessing approaches, and feature extraction approaches. Therefore, our research aimed to address the gap. This study aims to develop a machine learning model to detect the adulteration of Ethiopian injera. Since adulterated injera is illegal and producers are unwilling to give us the dataset, we have prepared 2880 adulterated and 960 pure injera image datasets ourselves by collecting red teff, white teff, sawdust, mold injera, and gypsum from the market and timber house within two seasons at different places. For our work, we use 5,760 augmented injera images. To achieve the objective, we applied an experimental research design. We have done experiments with two cases: binary classification and multiclass classification. In case of multiclass, first we experimented without applying image filtering and denoising techniques the accuracy was 54% for CNN and 57% for SVM classifiers. After applying preprocessing the result for SVM is given as 79% but the result for CNN is 74%. In case of binary class first, we experimented without applaying image filtering and denoising techniques the accuracy was 66% for CNN and 74% for SVM classifiers. On the other hand, when we applied Gausian and median filter, Canny detector CNN feature extractor the classification accuracy was 92% for CNN and 84% for SVM classifier. Second, we have done an experiment using Gausian and median filter, Canny detector YOLOV7 feature extractor, based on the experimental results, a better accuracy was 95% for CNN and 88% for SVM classifiers. Finally, we applied post-processing and get an optimal accuracy of 99% in case of binary class for CNN classifier, and 83% in case of multiclass for SVM classifier. The CNN model's accuracy shows a promising future in the effectiveness of Ethiopians injera classification. Keywords: YOLOV7, Injera, Adulteration detection ,machine learning, CNN ,SVM.