6DoF WIDER Face Annotations - img2pose

WIDER FACE is the most used publicly dataset for training face detection methods, containing around 400k faces across the train, validation, and test sets. We provide six degrees of freedom (6DoF) annotations for around 170k faces on the WIDER FACE train set and 34k faces on the validation set, that were used to train our recent work img2pose. These annotations were acquired through a combination of manual and automatic detected points, that were then converted to 6DoF for training and testing. More details about the annotations can be found in our paper. The WIDER FACE images can be download here. Our 6DoF annotations here. Code, models, and details on how to prepare the dataset, train, and test can be found here.