The 300w is a face dataset made up of 300 in-the-wild photographs taken inside and outdoors. It encompasses a wide range of identity, expression, lighting, stance, occlusion, and facial size. The photographs were obtained by searching for "party," "conference," "protests," "football," and "celebrities" on Google. The 300w database has a higher percentage of partially-occluded photos and covers more expressions than the standard "neutral" or "smile," such as "surprise" or "scream," when compared to other in-the-wild datasets. A semi-automatic approach was used to annotate images using the 68-point mark-up.