Dairy cattle raised in indoor housing are exposed to potential sources of skin injuries that include swelling, wounds and other visually detectable lesions. Unfortunately, current diagnosis relies on qualitative visual evaluation and requires a trained observer to conduct the assessment. Computer vision assessment can be adapted to screen and potential cattle injuries, allowing any evaluator with minimum injury detection experience to obtain rapid assessment of the severity of cow injury almost instantly. To implement and test such a system, a machine-learning tool was developed using a data set of 2,364 images of both injured and non-injured cows. Skin injuries corresponded to either broken hair (BH), complete hair loss (CHL), white or dry scabs (WDS), red and wet scabs (RWS) and open wounds (WO) and were present at 3 injury locations (knee, lateral calcanei, and lateral tarsal). The data set was split into 3 separate subsets each corresponding to one of the injury locations. Regions of interest (ROIs) were first extracted systematically from each image using a pattern recognition module and were then resized and normalized. ROIs were then passed to a convolutional neural network (CNN) classifier constructed for each injury type in each location. These image subsets were then split into a training set, a validation set, and a test set to allow CNN optimization and evaluation of system performance. The classification accuracy rates of the CNN show an average accuracy of 93.9, 68.1, 76.4, 74.6 and 98.5% accurately detected injuries for BH, CHL, WDS, RWS and WO respectively. A ZeroR classifier predicting the majority category class was used as a baseline, and average accuracy values for this model were 90.3, 61.4, 77.0, 72.2, and 98.7% for BH, CHL, WDS, RWS and WO respectively. The CNN classifier for BH, CHL, and RWS outperformed the ZeroR model and a model performing random class assignment. According to test results, this system can be used as a rapid cow injury screening tool. A more extensive data set is being processed to further increase CNN accuracy.