time cost

fa = face_alignment.FaceAlignment(face_alignment.LandmarksType._2D, flip_input=False, device='cuda')
fa.get_landmarks(img)

how much seconds is it cost?
I get 8 seconds for 3 pictures.

The issue asks about the execution time of `fa.get_landmarks(img)`. Start by reading the library's source code for the FaceAlignment class and the get_landmarks method to understand the pipeline. Create a small benchmark script to measure the time per image, considering factors like image size, GPU vs CPU, and model loading. Check if there are existing performance tests or profiling tools in the repository. The goal is to quantify the time cost and identify potential bottlenecks.

Rédigé par le modèle d'indexation à partir du texte de l'issue.