def __init__(self):
self.model_type = 'svm'
self.svm = SVM.Params()
self.xgb = XGB.Params()
self.mlp = MLP.Params()
self.help = {
'model_type':
'learning method used for decision making in the active state: '
'svm: Support Vector Machine,'
'xgb: XGBoost,'
'mlp: Multi Layer Perceptron,'
'cnn: Convolutional Neural Network',
'svm':
'parameters for the SVM module',
'xgb':
'parameters for the XGB module',
'mlp':
'parameters for the MLP module',
}
def __init__(self):
self.model_type = 'svm'
self.max_occlusion = 50
self.threshold_ratio = 0.6
self.threshold_dist = 3
self.overlap_box = 0.5
self.overlap_pos = 0.5
self.overlap_neg = 0.2
self.use_heuristic_features = 1
self.pause_for_debug = 0
self.verbose = 0
self.weight_tracking = 1
self.weight_association = 1
self.copy_while_learning = 1
self.svm = SVM.Params()
self.xgb = XGB.Params()
self.mlp = MLPParams()
self.help = {
'model_type': 'learning method used for decision making in the lost state: '
'svm: Support Vector Machine,'
'xgb: XGBoost,'
'mlp: Multi Layer Perceptron,'
'cnn: Convolutional Neural Network',
'max_occlusion': 'maximum number of consecutive frames for which the target is allowed to remain in '
'the Lost state before it transitions to inactive',
'threshold_ratio': 'aspect ratio threshold in target association - this is the minimum ratio between the'
' heights of the last known location and a candidate detection for the latter to be '
'considered a possible match',
'threshold_dist': ' distance threshold in target association, multiple of the width of target - '
'this is the maximum ratio of the Euclidean distance between the centers of the last '
'known target location and a candidate detection with the width of the former '
'for the latter to be considered a possible match',
'overlap_box': 'minimum overlap (IOU) between the LK result and the best matching detection for the final'
' object location to be computed as a weighted average of the two; if the overlap is'
' less than this, then the detection itself is used as this location;',
'overlap_pos': 'minimum overlap (IOU) of the ground truth location with the best matching detection and '
'the computed object location for the corresponding feature vector to be regarded '
'as a positive training sample for learning',
'overlap_neg': 'maximum overlap (IOU) of the ground truth location with the best matching detection or '
'the computed object location for the corresponding feature vector to be regarded '
'as a negative training sample for learning',
'use_heuristic_features': 'augment tracker/templates features with heuristics to '
'construct policy classifier features; '
'setting to 0 will use only tracker features',
'copy_while_learning': 'original code did not update templates state matrices while getting '
'learning features for some unknown annoying reason; toggle this behavior',
'pause_for_debug': 'pause execution for debugging',
'verbose': 'Enable printing of some general diagnostic messages',
'weight_tracking': 'weight given to the tracked location while computing the weighted average '
'of this and the best matching detection as the final object location during association',
'weight_association': 'weight given to the best matching detection while computing the weighted average '
'of this and the tracked location as the final object location during association',
'svm': 'parameters for the SVM module',
'xgb': 'parameters for the XGB module',
'mlp': 'parameters for the MLP module'
}
