Skip to content

Latest commit

 

History

History
155 lines (118 loc) · 4.55 KB

README.md

File metadata and controls

155 lines (118 loc) · 4.55 KB
Raw

Facial Features

Input

(Image from https://pixabay.com/photos/people-cowboy-male-hat-person-875597/)

Face Detector: BlazeFace

  • ailia input shape: (1, 3, 128, 128) RGB channel order
  • Pixel value range: [-1, 1]

Face Landmarks Estimator: Face Mesh

  • ailia input shape: (batch_size, 3, 192, 192) RGB channel order
  • Pixel value range: [-1, 1]

Eyelids Classifier: ax Eyelids

  • ailia input shape: (batch_size, 3, 256, 256) RGB channel order
  • Pixel value range: [0, 1] before normalization
  • Preprocessing: normalization using ImageNet statistics

Eyelashes Classifier: ax Eyelashes

  • ailia input shape: (batch_size, 3, 256, 256) RGB channel order
  • Pixel value range: [0, 1] before normalization
  • Preprocessing: normalization using ImageNet statistics

Facial Hair Classifier: ax Facial Hair

  • ailia input shape: (batch_size, 3, 256, 256) RGB channel order
  • Pixel value range: [0, 1] before normalization
  • Preprocessing: normalization using ImageNet statistics

Output

ROI
==============================================================
top_left = (-12.06, 139.10) (x, y)
top_right = (918.57, 50.24) (x, y)
bottom_left = (76.80, 1069.73) (x, y)
bottom_right = (1007.43, 980.87) (x, y)

Eyelids
==============================================================
class_count = 2
+ idx = 0
  category = 0 [double]
  prob = 0.998113751411438
+ idx = 1
  category = 1 [single]
  prob = 0.0018862121505662799

Eyelashes
==============================================================
class_count = 3
+ idx = 0
  category = 0 [dense]
  prob = 0.8559471368789673
+ idx = 1
  category = 1 [moderate]
  prob = 0.09105094522237778
+ idx = 2
  category = 2 [sparse]
  prob = 0.05300195515155792

Facial hair
==============================================================
label_count = 3
+ category = 0 [moustache]
  prob = 0.9992972612380981
+ category = 1 [beard]
  prob = 0.9995272755622864
+ category = 2 [mouth_side_hair]
  prob = 0.9998342990875244

Face Detector: BlazeFace

  • ailia Predict API output:
    • Bounding boxes and keypoints
      • Shape: (1, 896, 16)
    • Classification confidences
      • Shape: (1, 896, 1)
  • With helper functions, filtered detections with keypoints can be obtained.

Face Landmarks Estimator: Face Mesh

  • ailia Predict API output:
    • landmarks: 468 face landmarks with (x, y, z) coordinates
      • Shape: (1, batch_size*468*3)
      • x and y are in the range [0, 192] (to normalize, divide by the image width and height, 192). z represents the landmark depth with the depth at center of the head being the origin, and the smaller the value the closer the landmark is to the camera. The magnitude of z uses roughly the same scale as x.
    • confidences: raw confidence values. Applying the sigmoid function yields scores in [0, 1]
      • Shape: (1, batch_size*1)
  • With helper functions, image (original size) coordinates of eye centers, iris landmarks and cropped eye region image can be obtained.

Eyelids Classifier: ax Eyelids

  • ailia Predict API output:
    • eyelids: raw prediction values for eyelids classification. Applying the sigmoid function yields scores in [0, 1]
      • Shape: (batch_size, 2) [double, single] order

Eyelashes Classifier: ax Eyelashes

  • ailia Predict API output:
    • eyelashes: raw prediction values for eyelashes classification. Applying the sigmoid function yields scores in [0, 1]
      • Shape: (batch_size, 3) [dense, moderate, sparse] order

Facial Hair Classifier: ax Facial Hair

  • ailia Predict API output:
    • facial_hair: probability values for facial hair multi-label classification
      • Shape: (batch_size, 3) [moustache, beard, mouth_side_hair] order

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 ax_facial_features.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 ax_facial_features.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 ax_facial_features.py --video VIDEO_PATH --savepath SAVE_VIDEO_PATH

Additional options

  • -m or --mode: Allow to choose which classification(s) among [eyelids, eyelashes, facial_hair] to perform

Model Format

ONNX opset = 11