BEAT - tutorial/lbphs

This algorithm is a legacy one. The API has changed since its implementation. New versions and forks will need to be updated.

This algorithm is splittable

Endpoint Groups 1

Algorithms have at least one input and one output. All algorithm endpoints are organized in groups. Groups are used by the platform to indicate which inputs and outputs are synchronized together. The first group is automatically synchronized with the channel defined by the block in which the algorithm is deployed.

Unnamed group

Endpoint Name	Data Format	Nature
image	system/array_3d_uint8/1	Input
eye_centers	system/eye_positions/1	Input
features	system/array_1d_floats/1	Output

Parameters 8

Parameters allow users to change the configuration of an algorithm when scheduling an experiment

Name	Type	Default
crop-height	uint32	80
crop-width	uint32	64
right-eye-y	uint32	16
right-eye-x	uint32	15
left-eye-y	uint32	16
left-eye-x	uint32	48
block-size	uint32	12
block-overlap	uint32	11

import bob
import numpy

class Algorithm:

def __init__(self):
        self.crop_height = 80
        self.crop_width  = 64
        self.right_eye_y = 16
        self.right_eye_x = 15
        self.left_eye_y  = 16
        self.left_eye_x  = 48
        self.lbp_block_size = 8
        self.lbp_block_overlap = 0

def setup(self, parameters):
        self.crop_height = parameters.get('crop-height', self.crop_height)
        self.crop_width  = parameters.get('crop-width',  self.crop_width)
        self.right_eye_y = parameters.get('right-eye-y', self.right_eye_y)
        self.right_eye_x = parameters.get('right-eye-x', self.right_eye_x)
        self.left_eye_y  = parameters.get('left-eye-y',  self.left_eye_y)
        self.left_eye_x  = parameters.get('left-eye-x',  self.left_eye_x)

self.cropper = bob.ip.FaceEyesNorm(int(self.crop_height+4),
            int(self.crop_width+4), int(self.right_eye_y+2), int(self.right_eye_x+2),
            int(self.left_eye_y+2), int(self.left_eye_x+2))

self.lbp_block_size = parameters.get('block-size', self.lbp_block_size)
        self.blp_block_overlap = parameters.get('block-overlap',
            self.lbp_block_overlap)

self.lbp_extractor = bob.ip.LBPHSFeatures(
            int(self.lbp_block_size),
            int(self.lbp_block_size),
            int(self.lbp_block_overlap),
            int(self.lbp_block_overlap),
            bob.ip.LBP(8, 2., uniform=True)
        )

return True

def process(self, inputs, outputs):
        image = bob.ip.rgb_to_gray(inputs["image"].data.value)
        eye_centers = inputs["eye_centers"].data

cropped_image = self.cropper(image, float(eye_centers.right.y),
            float(eye_centers.right.x), float(eye_centers.left.y),
            float(eye_centers.left.x))

features = \
          numpy.hstack(self.lbp_extractor(cropped_image)).astype(numpy.float64)

outputs["features"].write({
          "value" : features
        })

return True

The code for this algorithm in Python
The ruler at 80 columns indicate suggested POSIX line breaks (for readability).
The editor will automatically enlarge to accomodate the entirety of your input
Use keyboard shortcuts for search/replace and faster editing. For example, use Ctrl-F (PC) or Cmd-F (Mac) to search through this box

This algorithms extracts Local Binary Pattern Histogram Sequence (LBPHS) features as introduced by [Ahonen04]. First, the image is aligned accorsing to the hand-labeled eye locations. Then, uniform circular Local Binary Patterns (LBPs) [Ojala96] are extracted from the aligned image. Afterwards, the image is split into (possibly overlapping) blocks, and a local histogram of LBP features is extracted for each block. Finally, all histograms are concatenated to form the full LBPHS feature vector.

[Ojala96]

Ojala, M. Pietikainen, D. Harwood: A comparative study of texture measures with classification based on feature distributions. Pattern Recognition 29 (1996) 51-59

[Ahonen04]

Ahonen, A. Hadid, M. Pietikainen: Face Recognition with Local Binary Patterns. ECCV (1) 2004: 469-481

Experiments

Attestation:

Privacy:

Status:

			Updated	Name	Databases/Protocols	Analyzers
				tutorial/tutorial/full_lbphs/1/atnt-lbphs-para	atnt/1@idiap_test_eyepos	tutorial/eerhter_postperf/1

This table shows the number of times this algorithm has been successfully run using the given environment. Note this does not provide sufficient information to evaluate if the algorithm will run when submitted to different conditions.

algorithms tutorial lbphs_features 2

Endpoint Groups 1

Unnamed group

Parameters 8

Experiments

algorithms

tutorial

lbphs_features

2