1#!/usr/bin/python3 2'''Copyright (c) 2021-2022 Amazon 3 4 Redistribution and use in source and binary forms, with or without 5 modification, are permitted provided that the following conditions 6 are met: 7 8 - Redistributions of source code must retain the above copyright 9 notice, this list of conditions and the following disclaimer. 10 11 - Redistributions in binary form must reproduce the above copyright 12 notice, this list of conditions and the following disclaimer in the 13 documentation and/or other materials provided with the distribution. 14 15 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 16 ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 17 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 18 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR 19 CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 20 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 21 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 22 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 23 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 24 NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 25 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26''' 27 28import numpy as np 29from tensorflow.keras.utils import Sequence 30 31class PLCLoader(Sequence): 32 def __init__(self, features, lost, nb_burg_features, batch_size): 33 self.batch_size = batch_size 34 self.nb_batches = features.shape[0]//self.batch_size 35 self.features = features[:self.nb_batches*self.batch_size, :, :] 36 self.lost = lost.astype('float') 37 self.lost = self.lost[:(len(self.lost)//features.shape[1]-1)*features.shape[1]] 38 self.nb_burg_features = nb_burg_features 39 self.on_epoch_end() 40 41 def on_epoch_end(self): 42 self.indices = np.arange(self.nb_batches*self.batch_size) 43 np.random.shuffle(self.indices) 44 offset = np.random.randint(0, high=self.features.shape[1]) 45 self.lost_offset = np.reshape(self.lost[offset:-self.features.shape[1]+offset], (-1, self.features.shape[1])) 46 self.lost_indices = np.random.randint(0, high=self.lost_offset.shape[0], size=self.nb_batches*self.batch_size) 47 48 def __getitem__(self, index): 49 features = self.features[self.indices[index*self.batch_size:(index+1)*self.batch_size], :, :] 50 burg_lost = (np.random.rand(features.shape[0], features.shape[1]) > .1).astype('float') 51 burg_lost = np.reshape(burg_lost, (features.shape[0], features.shape[1], 1)) 52 burg_mask = np.tile(burg_lost, (1,1,self.nb_burg_features)) 53 54 lost = self.lost_offset[self.lost_indices[index*self.batch_size:(index+1)*self.batch_size], :] 55 lost = np.reshape(lost, (features.shape[0], features.shape[1], 1)) 56 lost_mask = np.tile(lost, (1,1,features.shape[2])) 57 in_features = features*lost_mask 58 in_features[:,:,:self.nb_burg_features] = in_features[:,:,:self.nb_burg_features]*burg_mask 59 60 #For the first frame after a loss, we don't have valid features, but the Burg estimate is valid. 61 #in_features[:,1:,self.nb_burg_features:] = in_features[:,1:,self.nb_burg_features:]*lost_mask[:,:-1,self.nb_burg_features:] 62 out_lost = np.copy(lost) 63 #out_lost[:,1:,:] = out_lost[:,1:,:]*out_lost[:,:-1,:] 64 65 out_features = np.concatenate([features[:,:,self.nb_burg_features:], 1.-out_lost], axis=-1) 66 burg_sign = 2*burg_lost - 1 67 # last dim is 1 for received packet, 0 for lost packet, and -1 when just the Burg info is missing 68 inputs = [in_features*lost_mask, lost*burg_sign] 69 outputs = [out_features] 70 return (inputs, outputs) 71 72 def __len__(self): 73 return self.nb_batches 74