From 13c7964dbfa216cbe9868c421377790045dc5116 Mon Sep 17 00:00:00 2001
From: Olegs NIKISINS <onikisins@italix03.idiap.ch>
Date: Mon, 29 May 2017 13:31:57 +0200
Subject: [PATCH] Changed the LBP+SVM pipeline settings to partially match
Ivana's paper
---
.../face/algorithm/VideoSvmPadAlgorithm.py | 81 +++++++++++++++++--
.../config/extractor/video_lbp_histogram.py | 25 ++----
bob/pad/face/config/grid.py | 4 +-
.../config/preprocessor/video_face_crop.py | 14 ++--
setup.py | 5 +-
5 files changed, 92 insertions(+), 37 deletions(-)
diff --git a/bob/pad/face/algorithm/VideoSvmPadAlgorithm.py b/bob/pad/face/algorithm/VideoSvmPadAlgorithm.py
index a6225ecb..ef921338 100644
--- a/bob/pad/face/algorithm/VideoSvmPadAlgorithm.py
+++ b/bob/pad/face/algorithm/VideoSvmPadAlgorithm.py
@@ -17,6 +17,10 @@ import numpy as np
import bob.learn.libsvm
+import bob.io.base
+
+import os
+
#==============================================================================
# Main body :
@@ -52,6 +56,10 @@ class VideoSvmPadAlgorithm(Algorithm):
``mean_std_norm_flag`` : :py:class:`bool`
Perform mean-std normalization of data if set to True. Default: False.
+
+ ``frame_level_scores`` : :py:class:`bool`
+ Return scores for each frame individually if True. Otherwise, return a
+ single score per video. Default: False.
"""
def __init__(self,
@@ -59,7 +67,8 @@ class VideoSvmPadAlgorithm(Algorithm):
kernel_type = 'RBF',
n_samples = 10000,
trainer_grid_search_params = { 'cost': [2**p for p in range(-5, 16, 2)], 'gamma': [2**p for p in range(-15, 4, 2)]},
- mean_std_norm_flag = False):
+ mean_std_norm_flag = False,
+ frame_level_scores = False):
Algorithm.__init__(self,
@@ -68,6 +77,7 @@ class VideoSvmPadAlgorithm(Algorithm):
n_samples = n_samples,
trainer_grid_search_params = trainer_grid_search_params,
mean_std_norm_flag = mean_std_norm_flag,
+ frame_level_scores = frame_level_scores,
performs_projection=True,
requires_projector_training=True)
@@ -76,6 +86,7 @@ class VideoSvmPadAlgorithm(Algorithm):
self.n_samples = n_samples
self.trainer_grid_search_params = trainer_grid_search_params
self.mean_std_norm_flag = mean_std_norm_flag
+ self.frame_level_scores = frame_level_scores
self.machine = None
@@ -434,7 +445,8 @@ class VideoSvmPadAlgorithm(Algorithm):
def train_svm(self, training_features, n_samples = 10000,
machine_type = 'C_SVC', kernel_type = 'RBF',
trainer_grid_search_params = { 'cost': [2**p for p in range(-5, 16, 2)], 'gamma': [2**p for p in range(-15, 4, 2)]},
- mean_std_norm_flag = False):
+ mean_std_norm_flag = False,
+ projector_file = ""):
"""
First, this function tunes the hyper-parameters of the SVM classifier using
grid search on the sub-sets of training data. Train and cross-validation
@@ -469,6 +481,12 @@ class VideoSvmPadAlgorithm(Algorithm):
``mean_std_norm_flag`` : :py:class:`bool`
Perform mean-std normalization of data if set to True. Default: False.
+ ``projector_file`` : :py:class:`str`
+ The name of the file to save the trained projector to. Only the path
+ of this file is used in this function. The file debug_data.hdf5 will
+ be save in this path. This file contains information, which might be
+ usefull for debugging.
+
**Returns:**
``machine`` : object
@@ -525,6 +543,18 @@ class VideoSvmPadAlgorithm(Algorithm):
setattr(trainer, key, selected_params[key]) # set the params of trainer
+ # Save the data, which is usefull for debugging.
+ debug_file = os.path.join( os.path.split(projector_file)[0], "debug_data.hdf5" )
+ debug_dict = {}
+ debug_dict['precisions_train'] = precisions_train
+ debug_dict['precisions_cv'] = precisions_cv
+ debug_dict['cost'] = selected_params['cost']
+ debug_dict['gamma'] = selected_params['gamma']
+ f = bob.io.base.HDF5File(debug_file, 'w') # open hdf5 file to save the debug data
+ for key in debug_dict.keys():
+ f.set(key, debug_dict[key])
+ del f
+
# training_features[0] - training features for the REAL class.
real = self.convert_list_of_frame_cont_to_array(training_features[0]) # output is array
# training_features[1] - training features for the ATTACK class.
@@ -574,10 +604,14 @@ class VideoSvmPadAlgorithm(Algorithm):
machine_type = self.machine_type,
kernel_type = self.kernel_type,
trainer_grid_search_params = self.trainer_grid_search_params,
- mean_std_norm_flag = self.mean_std_norm_flag)
+ mean_std_norm_flag = self.mean_std_norm_flag,
+ projector_file = projector_file)
+
+ f = bob.io.base.HDF5File(projector_file, 'w') # open hdf5 file to save to
+ machine.save(f) # save the machine and normalization parameters
- machine.save(projector_file)
+ del f
#==========================================================================
@@ -596,7 +630,11 @@ class VideoSvmPadAlgorithm(Algorithm):
The file to read the projector from.
"""
- self.machine = bob.learn.libsvm.Machine(projector_file)
+ f = bob.io.base.HDF5File(projector_file, 'a')
+
+ self.machine = bob.learn.libsvm.Machine(f)
+
+ del f
#==========================================================================
@@ -648,22 +686,49 @@ class VideoSvmPadAlgorithm(Algorithm):
**Returns:**
``score`` : :py:class:`float`
+ or a list of scores containing individual score for each frame.
A score value for the object ``toscore``.
A probability of a sample being a real class.
"""
- score = np.mean(toscore, axis=0)[0]
+ if self.frame_level_scores:
+
+ score = toscore[:,0] # here score is a list containing scores for each frame
+
+ else:
+
+ score = np.mean(toscore, axis=0)[0] # compute a single score per video
return score
#==========================================================================
def score_for_multiple_projections(self, toscore):
- """Returns the difference between log likelihoods of being real or attack"""
+ """
+ Returns a list of scores computed by the score method of this class.
+
+ **Parameters:**
+
+ ``toscore`` : 2D :py:class:`numpy.ndarray`
+ An array containing scores computed by score() method of this class.
+
+ **Returns:**
+
+ ``list_of_scores`` : list
+ A list containing the scores.
+ """
# import ipdb; ipdb.set_trace()
- return [self.score(toscore)]
+ if self.frame_level_scores:
+
+ list_of_scores = self.score(toscore)
+
+ else:
+
+ list_of_scores = [self.score(toscore)]
+
+ return list_of_scores
diff --git a/bob/pad/face/config/extractor/video_lbp_histogram.py b/bob/pad/face/config/extractor/video_lbp_histogram.py
index d9d3e96c..ee0ec7f7 100644
--- a/bob/pad/face/config/extractor/video_lbp_histogram.py
+++ b/bob/pad/face/config/extractor/video_lbp_histogram.py
@@ -6,25 +6,16 @@ from bob.pad.face.extractor import VideoLBPHistogram
#=======================================================================================
# Define instances here:
-lbptype='regular'
+lbptype='uniform'
elbptype='regular'
-rad=3
+rad=1
neighbors=8
circ=False
dtype=None
-video_lbp_histogram_extractor_n8r3 = VideoLBPHistogram(lbptype=lbptype,
- elbptype=elbptype,
- rad=rad,
- neighbors=neighbors,
- circ=circ,
- dtype=dtype)
-
-lbptype='uniform'
-
-video_lbp_histogram_extractor_n8r3_uniform = VideoLBPHistogram(lbptype=lbptype,
- elbptype=elbptype,
- rad=rad,
- neighbors=neighbors,
- circ=circ,
- dtype=dtype)
\ No newline at end of file
+video_lbp_histogram_extractor_n8r1_uniform = VideoLBPHistogram(lbptype=lbptype,
+ elbptype=elbptype,
+ rad=rad,
+ neighbors=neighbors,
+ circ=circ,
+ dtype=dtype)
diff --git a/bob/pad/face/config/grid.py b/bob/pad/face/config/grid.py
index 3cddf7e5..12d4c405 100644
--- a/bob/pad/face/config/grid.py
+++ b/bob/pad/face/config/grid.py
@@ -19,7 +19,7 @@ idiap = Grid(
number_of_enrollment_jobs=32,
enrollment_queue='8G-io-big',
- number_of_scoring_jobs=50,
+ number_of_scoring_jobs=1,
scoring_queue='8G-io-big',
)
@@ -39,6 +39,6 @@ idiap_user_machines = Grid(
number_of_enrollment_jobs=32,
enrollment_queue='8G',
- number_of_scoring_jobs=50,
+ number_of_scoring_jobs=1,
scoring_queue='8G',
)
diff --git a/bob/pad/face/config/preprocessor/video_face_crop.py b/bob/pad/face/config/preprocessor/video_face_crop.py
index b210d189..61eaf725 100644
--- a/bob/pad/face/config/preprocessor/video_face_crop.py
+++ b/bob/pad/face/config/preprocessor/video_face_crop.py
@@ -6,15 +6,15 @@ from bob.pad.face.preprocessor import VideoFaceCrop
#=======================================================================================
# Define instances here:
-cropped_image_size = (100, 100) # The size of the resulting face
-cropped_positions = { 'topleft' : (0,0) , 'bottomright' : cropped_image_size}
+cropped_image_size = (64, 64) # The size of the resulting face
+cropped_positions = {'topleft' : (0,0) , 'bottomright' : cropped_image_size}
fixed_positions = None
-mask_sigma = None # The sigma for random values areas outside image
-mask_neighbors = 5 # The number of neighbors to consider while extrapolating
-mask_seed = None # The seed for generating random values during extrapolation
-color_channel = 'gray' # Convert image to gray-scale format
+mask_sigma = None # The sigma for random values areas outside image
+mask_neighbors = 5 # The number of neighbors to consider while extrapolating
+mask_seed = None # The seed for generating random values during extrapolation
+color_channel = 'gray' # Convert image to gray-scale format
-video_face_crop_preproc_100_100 = VideoFaceCrop(cropped_image_size = cropped_image_size,
+video_face_crop_preproc_64_64 = VideoFaceCrop(cropped_image_size = cropped_image_size,
cropped_positions = cropped_positions,
fixed_positions = fixed_positions,
mask_sigma = mask_sigma,
diff --git a/setup.py b/setup.py
index 53dab1e9..791a68e8 100644
--- a/setup.py
+++ b/setup.py
@@ -100,13 +100,12 @@ setup(
# registered preprocessors:
'bob.pad.preprocessor': [
- 'video-face-crop-preproc-100 = bob.pad.face.config.preprocessor.video_face_crop:video_face_crop_preproc_100_100',
+ 'video-face-crop-preproc-64 = bob.pad.face.config.preprocessor.video_face_crop:video_face_crop_preproc_64_64',
],
# registered preprocessors:
'bob.pad.extractor': [
- 'video-lbp-histogram-extractor-n8r3 = bob.pad.face.config.extractor.video_lbp_histogram:video_lbp_histogram_extractor_n8r3',
- 'video-lbp-histogram-extractor-n8r3-uniform = bob.pad.face.config.extractor.video_lbp_histogram:video_lbp_histogram_extractor_n8r3_uniform',
+ 'video-lbp-histogram-extractor-n8r1-uniform = bob.pad.face.config.extractor.video_lbp_histogram:video_lbp_histogram_extractor_n8r1_uniform',
],
# registered algorithms:
--
GitLab