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.
| Endpoint Name | Data Format | Nature |
|---|---|---|
| image | system/array_2d_uint8/1 | Input |
| features | system/array_2d_floats/1 | Output |
Parameters allow users to change the configuration of an algorithm when scheduling an experiment
| Name | Description | Type | Default | Range/Choices |
|---|---|---|---|---|
| regions_cols | Number of regions in column | uint8 | 8 | [1, 100] |
| lbp_maxradius | Maxmum radius of milti LBP | uint8 | 1 | [1, 10] |
| regions_rows | Number of regions in row | uint8 | 8 | [1, 100] |
| lbp_minradius | Minimum radius of milti LBP | uint8 | 1 | [1, 10] |
| lbp_circular | LBP Circular | bool | True | |
| lbp_uniform | LBP Uniform | bool | True | |
| lbp_neighbor_count | LBP Neighbour count | uint8 | 8 | [8, 32] |
xxxxxxxxxximport bobimport numpy as npclass Algorithm: def __init__(self): self.lbp_minradius = 1 self.lbp_maxradius = 10 self.lbp_neighbor_count = 8 self.lbp_uniform = True self.lbp_circular = True self.regions_cols = 8 self.regions_rows = 8 def setup(self, parameters): self.lbp_minradius = parameters.get('lbp_minradius', self.lbp_minradius) self.lbp_maxradius = parameters.get('lbp_maxradius', self.lbp_maxradius) self.lbp_neighbor_count = parameters.get('lbp_neighbor_count', self.lbp_neighbor_count) self.lbp_uniform = parameters.get('lbp_uniform', self.lbp_uniform) self.lbp_circular = parameters.get('lbp_circular', self.lbp_circular) self.regions_cols = parameters.get('regions_cols', self.regions_cols) self.regions_rows = parameters.get('regions_rows', self.regions_rows) self.lbp_rotation_invariant = False self.lbp_compare_to_average = False self.lbp_add_average = False self.lbp_extractors = [bob.ip.LBP(np.int(self.lbp_neighbor_count), np.float(r), self.lbp_circular, False, False, self.lbp_uniform, False) for r in xrange(self.lbp_minradius, self.lbp_maxradius+1)] return True def process(self, inputs, outputs): cropped_image = inputs["image"].data.value regions_width = (cropped_image.shape[1]-2*self.lbp_maxradius)/self.regions_cols regions_height = (cropped_image.shape[0]-2*self.lbp_maxradius)/self.regions_rows features = np.empty([self.regions_cols*self.regions_rows,0]) assert(features.shape[0]!=self.regions_cols*self.regions_rows,'features.shape[0]!=self.regions_cols*self.regions_rows') std_lbp_output_image = np.ndarray(self.lbp_extractors[self.lbp_maxradius-1].get_lbp_shape(cropped_image), dtype = np.uint16 ) for i in xrange(len(self.lbp_extractors)): lbp_output_image = np.ndarray(self.lbp_extractors[i].get_lbp_shape(cropped_image), dtype = np.uint16 ) self.lbp_extractors[i](cropped_image,lbp_output_image) lbp_output_image = bob.ip.crop(lbp_output_image, int(self.lbp_maxradius-self.lbp_extractors[i].radius), int(self.lbp_maxradius-self.lbp_extractors[i].radius), std_lbp_output_image.shape[0],std_lbp_output_image.shape[1]) lbp_img_blks = bob.ip.block(lbp_output_image,regions_height,regions_width,0,0) lbp_img_blks = lbp_img_blks.reshape(lbp_img_blks.shape[0]*lbp_img_blks.shape[1],lbp_img_blks.shape[2],lbp_img_blks.shape[3]) lbp_shists = np.empty([self.regions_cols*self.regions_rows,self.lbp_extractors[i].max_label]) for j in xrange(lbp_img_blks.shape[0]): lbp_hist = np.float32(bob.ip.histogram(lbp_img_blks[j],self.lbp_extractors[i].max_label-1)) lbp_hist /= lbp_hist.sum() lbp_shists[j] = lbp_hist features = np.hstack((features,lbp_shists))# if features.shape[1]==118:# print a# if features.shape[0]==64:# print b assert(features.shape[1]!=59,'features.shape[1]!=59') assert(features.shape[1]!=self.lbp_extractors[i].max_label*len(self.lbp_extractors),'number of radii not match')# assert(features.shape[0]!=64,'features.shape[0]!=64')# assert(features.shape[1]!=59,'features.shape[1]!=59') outputs["features"].write({ "value" : features }) return TrueThe 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
Refer to: Chi-Ho Chan, Josef Kittler, Kieron Messer: Multi-scale Local Binary Pattern Histograms for Face Recognition. 809-818
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.