How to speed up the deserialization function in dlib

I am using the C ++ version of dlib to extract the positions of facial cues in an image. I would like to shorten the lead time. It now takes about 2.5 seconds to execute the executed file and determine the landmark in one image.

After analysis, I found that the main bottleneck is the deserialization function, which takes the path to the shape predictor and loads that file into the shape predictor object. This function takes 1.5 seconds.

Below is a code snippet.

shape_predictor sp;
deserialize("shape_predictor_68_face_landmarks.dat") >> sp; 

      

I wonder if there might be a way to speed up the deserialization function.

+3
c ++ performance gcc computer-vision dlib


source to share


No one has answered this question yet

Check out similar questions:

3295
Why is the Android emulator so slow? How can we speed up the development of an Android emulator?
2302
How does database indexing work?
1248
How can you speed up Eclipse?
3
How to get the coordinates of the position of points in the program for determining the landmark of the face dlib?
2
Deserializing shape_predictor_68_face_landmarks.dat only once for real time face land orientation detection in android studio
1
python: Unexpected version found when deserializing dlib :: shape_predictor
1
Android JNI - load dlib form prediction file
1
Converting tagged faces to xml - dlib
0
Dlib get_frontal_face_detector not responding
0
Get ROI from cv2 dlib face landmarks



All Articles
Loading...
X
Show
Funny
Dev
Pics