Ensemble models

Author: pratikgujjar

code/autoencoder_model/scripts/config_sigc.py

@@ -70,7 +70,7 @@
 
 PRINT_MODEL_SUMMARY = True
 SAVE_MODEL = True
-PLOT_MODEL = False
+PLOT_MODEL = True
 SAVE_GENERATED_IMAGES = True
 SHUFFLE = True
 VIDEO_LENGTH = 16
@@ -82,7 +82,7 @@
 A_TRAIN_RATIO = 1
 C_TRAIN_RATIO = 1
 RAM_DECIMATE = True
-RETRAIN_CLASSIFIER = True
+RETRAIN_CLASSIFIER = False
 CLASS_TARGET_INDEX = 8
 ROT_MAX = 5
 SFT_H_MAX = 0.02
@@ -101,8 +101,8 @@
 print ("Loading network/training configuration...")
 print ("Config file: " + str(__name__))
 
-BATCH_SIZE = 25
-NB_EPOCHS_CLASS = 100
+BATCH_SIZE = 10
+NB_EPOCHS_CLASS = 30
 
 OPTIM_C = Adam(lr=0.0000002, beta_1=0.5)
 # OPTIM_C = SGD(lr=0.0001, momentum=0.9, nesterov=True)
@@ -125,9 +125,9 @@
 
 def schedule(epoch_idx):
     if (epoch_idx + 1) < lr_schedule[0]:
-        return 0.0001
+        return 0.00001
     elif (epoch_idx + 1) < lr_schedule[1]:
-        return 0.00001  # lr_decay_ratio = 10
+        return 0.000001  # lr_decay_ratio = 10
     elif (epoch_idx + 1) < lr_schedule[2]:
-        return 0.000001
-    return 0.000001
+        return 0.0000001
+    return 0.0000001

code/autoencoder_model/scripts/config_softc.py

@@ -0,0 +1,133 @@
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from keras.optimizers import SGD
+from keras.optimizers import Adam
+from keras.optimizers import RMSprop
+from keras import backend as K
+K.set_image_dim_ordering('tf')
+import socket
+import os
+
+# -------------------------------------------------
+# Background config:
+hostname = socket.gethostname()
+if hostname == 'baymax':
+    path_var = 'baymax/'
+elif hostname == 'walle':
+    path_var = 'walle/'
+elif hostname == 'bender':
+    path_var = 'bender/'
+else:
+    path_var = 'zhora/'
+
+DATA_DIR= '/local_home/JAAD_Dataset/iros/resized_imgs_208_sorted/train/'
+# DATA_DIR= '/local_home/data/KITTI_data/'
+
+TEST_DATA_DIR= '/local_home/JAAD_Dataset/iros/resized_imgs_208_sorted/test/'
+
+VAL_DATA_DIR= '/local_home/JAAD_Dataset/iros/resized_imgs_208_sorted/val/'
+
+PRETRAINED_C3D= '/home/pratik/git_projects/c3d-keras/models/sports1M_weights_tf.json'
+PRETRAINED_C3D_WEIGHTS= '/home/pratik/git_projects/c3d-keras/models/sports1M_weights_tf.h5'
+
+MODEL_DIR = './../' + path_var + 'models'
+if not os.path.exists(MODEL_DIR):
+    os.mkdir(MODEL_DIR)
+
+CHECKPOINT_DIR = './../' + path_var + 'checkpoints'
+if not os.path.exists(CHECKPOINT_DIR):
+    os.mkdir(CHECKPOINT_DIR)
+
+GEN_IMAGES_DIR = './../' + path_var + 'generated_images'
+if not os.path.exists(GEN_IMAGES_DIR):
+    os.mkdir(GEN_IMAGES_DIR)
+
+CLA_GEN_IMAGES_DIR = GEN_IMAGES_DIR + '/cla_gen/'
+if not os.path.exists(CLA_GEN_IMAGES_DIR):
+    os.mkdir(CLA_GEN_IMAGES_DIR)
+
+# ATTN_WEIGHTS_DIR = './../' + path_var + 'attn_weights'
+# if not os.path.exists(ATTN_WEIGHTS_DIR):
+#     os.mkdir(ATTN_WEIGHTS_DIR)
+
+LOG_DIR = './../' + path_var + 'logs'
+if not os.path.exists(LOG_DIR):
+    os.mkdir(LOG_DIR)
+
+TF_LOG_DIR = './../' + path_var + 'tf_logs'
+if not os.path.exists(TF_LOG_DIR):
+    os.mkdir(TF_LOG_DIR)
+
+TF_LOG_CLA_DIR = './../' + path_var + 'tf_cla_logs'
+if not os.path.exists(TF_LOG_CLA_DIR):
+    os.mkdir(TF_LOG_CLA_DIR)
+
+TEST_RESULTS_DIR = './../' + path_var + 'test_results'
+if not os.path.exists(TEST_RESULTS_DIR):
+    os.mkdir(TEST_RESULTS_DIR)
+
+PRINT_MODEL_SUMMARY = True
+SAVE_MODEL = True
+PLOT_MODEL = False
+SAVE_GENERATED_IMAGES = True
+SHUFFLE = True
+VIDEO_LENGTH = 16
+IMG_SIZE = (128, 208, 3)
+VIS_ATTN = True
+CLASSIFIER = True
+BUF_SIZE = 10
+LOSS_WEIGHTS = [1, 1]
+A_TRAIN_RATIO = 1
+C_TRAIN_RATIO = 1
+RAM_DECIMATE = True
+RETRAIN_CLASSIFIER = True
+CLASS_TARGET_INDEX = 8
+ROT_MAX = 5
+SFT_H_MAX = 0.02
+SFT_V_MAX = 0.02
+ZOOM_MAX = 0.2
+BRIGHT_RANGE_L = 0.5
+BRIGHT_RANGE_H = 1.5
+
+ped_actions = ['slow down', 'standing', 'walking', 'speed up', 'nod', 'unknown',
+               'clear path', 'handwave', 'crossing', 'looking', 'no ped']
+
+simple_ped_set = ['standing', 'crossing', 'no ped']
+
+# -------------------------------------------------
+# Network configuration:
+print ("Loading network/training configuration...")
+print ("Config file: " + str(__name__))
+
+BATCH_SIZE = 25
+NB_EPOCHS_CLASS = 100
+
+OPTIM_C = Adam(lr=0.0000002, beta_1=0.5)
+# OPTIM_C = SGD(lr=0.0001, momentum=0.9, nesterov=True)
+# OPTIM_C = RMSprop(lr=0.0001, rho=0.9)
+
+# lr_schedule = [10, 20, 30]  # epoch_step
+
+# def schedule(epoch_idx):
+#     if (epoch_idx + 1) < lr_schedule[0]:
+#         return 0.00000001
+#     elif (epoch_idx + 1) < lr_schedule[1]:
+#         return 0.000000001  # lr_decay_ratio = 10
+#     elif (epoch_idx + 1) < lr_schedule[2]:
+#         return 0.000000001
+#     return 0.000000001
+
+
+lr_schedule = [7, 15, 22]  # epoch_step
+
+
+def schedule(epoch_idx):
+    if (epoch_idx + 1) < lr_schedule[0]:
+        return 0.00001
+    elif (epoch_idx + 1) < lr_schedule[1]:
+        return 0.000001  # lr_decay_ratio = 10
+    elif (epoch_idx + 1) < lr_schedule[2]:
+        return 0.0000001
+    return 0.000001