The document describes a deep neural network-based system for detecting adulteration in dal grains. It involves developing a convolutional neural network (CNN) model to classify dal grains into types like kesar and toor based on texture and color features extracted from images. The proposed system architecture involves pre-processing input images, extracting features using CNN, classifying the dal type using the CNN and fully connected layers, and recognizing the dal grain based on the highest predictive value from a YOLO convolutional network. Key steps include image dataset collection, pre-processing, feature extraction, classification, and recognition to detect adulteration in dal grains from images.