In the process of preprocessing LRS2, this script generates mouth Regions of Interest (ROIs) in ${lrs2}/video. It shards all the videos listed in ${lrs2}/file.list into ${nshard} portions and generates mouth ROIs for the ${rank}-th shard, where "rank" is an integer in the range [0, nshard-1]. Face detection and landmark prediction are performed using the Dlib library.
The links to download the cnn_detector, face_detector, and mean_face can be found in the help message. However, it's not clear where to download the "mean_face," and further guidance is needed.
Exgc
commented
10 months ago
In the process of preprocessing LRS2, this script generates mouth Regions of Interest (ROIs) in ${lrs2}/video. It shards all the videos listed in ${lrs2}/file.list into ${nshard} portions and generates mouth ROIs for the ${rank}-th shard, where "rank" is an integer in the range [0, nshard-1]. Face detection and landmark prediction are performed using the Dlib library.
The links to download the cnn_detector, face_detector, and mean_face can be found in the help message. However, it's not clear where to download the "mean_face," and further guidance is needed.