Metal casting defects frequently occur in the manufacturing industry. There exist numerous types of defects, each capable of adversely affecting the entire industry. Featured-based transfer learning has demonstrated potential in metal casting classification. In this study, we investigate the ability of different feature-based transfer learning pipelines in classifying casting defects. A total dataset of 400 images containing two class types, defective and normal was utilized in a ratio of 70:15:15 for training, validation and testing. The VGG16 pre-trained convolutional neural network model was used to extract the features from the images. Different classifiers, namely Support Vector Machine (SVM), k-Nearest Neighbours (kNN) and Logistic Regression (LR) were utilized to classify images based on the extracted features. The present study implies that VGG16 + LR was able to distinguish the classes well with a testing accuracy of 98%. The study paves the way to the possible implementation of the suggested pipelines in the classification of casting defects.