Implementing computer vision with pretrained models