Callbacks
DrawInferences
paz.optimization.callbacks.DrawInferences(save_path, images, pipeline, topic='image', verbose=1)
Saves an image with its corresponding inferences
Arguments
- save_path: String. Path in which the images will be saved.
- images: List of numpy arrays of shape.
- pipeline: Function that takes as input an element of ''images'' and outputs a ''Dict'' with inferences.
- topic: Key to the ''inferences'' dictionary containing as value the drawn inferences.
- verbose: Integer. If is bigger than 1 messages would be displayed.
LearningRateScheduler
paz.optimization.callbacks.LearningRateScheduler(learning_rate, gamma_decay, scheduled_epochs, verbose=1)
Callback for reducing learning rate at specific epochs.
Arguments
- learning_rate: float. Indicates the starting learning rate.
- gamma_decay: float. In an scheduled epoch the learning rate is multiplied by this factor.
- scheduled_epochs: List of integers. Indicates in which epochs the learning rate will be multiplied by the gamma decay factor.
- verbose: Integer. If is bigger than 1 messages would be displayed.
EvaluateMAP
paz.optimization.callbacks.EvaluateMAP(data_manager, detector, period, save_path, iou_thresh=0.5)
Evaluates mean average precision (MAP) of an object detector.
Arguments
- data_manager: Data manager and loader class. See ''paz.datasets'' for examples.
- detector: Tensorflow-Keras model.
- period: Int. Indicates how often the evaluation is performed.
- save_path: Str.
- iou_thresh: Float.