300-VW stands for 300 Videos in the Wild, which is a video dataset for evaluating facial feature point tracking algorithms. The dataset contains 114 videos with a total of 218,595 frames, of which 64 videos are used for testing and 50 videos are used for training. In addition, all frames are annotated with the annotations (i.e. facial feature points) of the 300 W Challenge (a total of 68 facial feature points).