BEAT - tpereira/lbp

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

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
histogram	system/array_1d_floats/1	Output

Parameters 10

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

Name	Type	Default
block_size	int32	12
block_overlap	int32	0
cropped_image_height	uint32	80
cropped_image_width	uint32	64
right_eye_pos_x	uint32	16
right_eye_pos_y	uint32	15
left_eye_pos_x	uint32	16
left_eye_pos_y	uint32	48
lbp_P	int32	8
lbp_R	float64	2.0

import bob
import numpy

"""
Receive an image and computes the LBP histogram

LBP always squared

"""

class Algorithm:

def __init__(self):
    self.block_size           = 12
    self.block_overlap        = 0
    self.cropped_image_height = 80
    self.cropped_image_width  = 64
    self.right_eye_pos_x      = 16
    self.right_eye_pos_y      = 15
    self.left_eye_pos_x       = 16
    self.left_eye_pos_y       = 48
    self.lbp_P                = 8
    self.lbp_R                = 1.

def setup(self, parameters):
    self.block_size           = parameters.get('block_size', self.block_size)
    self.block_overlap        = parameters.get('block_overlap', self.block_overlap)
    self.cropped_image_height = parameters.get('cropped_image_height', self.cropped_image_height)
    self.cropped_image_width  = parameters.get('cropped_image_width', self.cropped_image_width)
    self.right_eye_pos_x      = parameters.get('right_eye_pos_x', self.right_eye_pos_x)
    self.right_eye_pos_y      = parameters.get('right_eye_pos_y', self.right_eye_pos_y)
    self.left_eye_pos_x       = parameters.get('left_eye_pos_x', self.left_eye_pos_x)
    self.left_eye_pos_y       = parameters.get('left_eye_pos_y', self.left_eye_pos_y)
    self.lbp_P                = parameters.get('lbp_P', self.lbp_P)
    self.lbp_R                = parameters.get('lbp_R', self.lbp_R)

self.face_norm = bob.ip.FaceEyesNorm(
          self.cropped_image_height, # cropped image height
          self.cropped_image_width, # cropped image width
          self.right_eye_pos_y, # Y of first position (usually: right eye)
          self.right_eye_pos_x, # X of first position (usually: right eye)
          self.left_eye_pos_y,  # Y of second position (usually: left eye)
          self.left_eye_pos_x   # X of second position (usually: left eye)
    )
   
    self.lbp = bob.ip.LBPHSFeatures(self.block_size, 
                                    self.block_size, 
                                    self.block_overlap, 
                                    self.block_overlap, 
                                    bob.ip.LBP(self.lbp_P, self.lbp_R, uniform = True)
                                   )

return True

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

eye_positions = inputs["eye_centers"].data

annotation_right_eye_y = int(eye_positions.right.y)
    annotation_right_eye_x = int(eye_positions.right.x)

annotation_left_eye_y = int(eye_positions.left.y)
    annotation_left_eye_x = int(eye_positions.left.x)

normalize_face = self.face_norm(image, annotation_right_eye_y, annotation_right_eye_x, annotation_left_eye_y, annotation_left_eye_x)
    histogram      = numpy.hstack(self.lbp(normalize_face)).astype(numpy.float64)
    #import ipdb; ipdb.set_trace();

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

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 algorithm computes:

A face crop, given the eyes center coordinates.
LBP histogram for each block (patch). The histograms for each block are then concatenated.

Details about LBP operator can be found in [Ojala02]

The inputs are:

image: RGB image
eye_centers: The coordinates of the eyes.

The output are the concatenated histograms of each block.

[Ojala02]

Ojala, M. Pietikainen, T. Maenpaa: Multiresolution gray-scale and rotation invariant texture classification with Local Binary Patterns. IEEE Transactions on Pattern Analysis and Machine inteligence 2002

No experiments are using this algorithm.

Scientific Python 2.7 (0.0.3) 8

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 tpereira lbp_extractor 1

Endpoint Groups 1

Unnamed group

Parameters 10

algorithms

tpereira

lbp_extractor

1