justinpinkney/trailer-faces-hq

Trailer Faces HQ (TFHQ)

Trailer Faces High Quality (TFHQ) is a large dataset of high-resolution face images sourced from movie trailers.

Details

TFHQ was collected by downloading all movie trailers and featurettes listed on the Apple Movie Trailers website as of August 2022. These 15,379 trailers were downloaded at Full HD (1080p) resolution, amounting to approximately 2 TB/507 hours of video.

Face detection was performed on every frame using the pre-trained Yolov5-face large model and any faces with bounding box less than 256 px in height, or confidence less than 0.5 were discarded. As film sequences contain many similar frames and these are often motion blurred due to the low frame rate of movies, filtering was performed to select the sharpest of similar consecutive frames, based on image similarity measured using a pre-trained CLIP ViT-B/32 [33] and sharpness measured using the variance of Laplacian measure. Finally the images are aligned following the same procedure as used for FFHQ.

As many of the final images exhibit undesirable properties such as occlusion, non-photographic faces, or overlaid text, we further perform quality filtering by training a classifier on several hundred subjectively determined “good”/“bad” example images, and use this to exclude predicted “bad” images leaving a total of 186,553 images in the dataset. Images are saved as quality=95 jpg files, filenames are 8 digit zero padded numbers.

Cite as:

@misc{pinkney2023tfhq,
      author = {Pinkney, Justin N. M.},
      title = {Trailer Faces HQ},
      year={2023},
      howpublished= {\url{https://huggingface.co/datasets/justinpinkney/trailer-faces-hq/}}
} 

Further samples