Code Compare Panel

Spot differences hiding in your code
Created on September 8|Last edited on September 9
Comment
﻿
﻿
 best traffic acc (50% data))
best overall IOU (20% data)
Expand 119 lines ...
config.framework = "fast.ai"    # AI framework used (for when we create other versions)config.framework = "fast.ai"    # AI framework used (for when we create other versions)
config.img_size = (360, 640)    # dimensions of resized image - can be 1 dim or tupleconfig.img_size = (360, 640)    # dimensions of resized image - can be 1 dim or tuple
-config.batch_size = 8           # Batch size during training+config.batch_size = 7           # Batch size during training
config.epochs = 10            # Number of epochs for trainingconfig.epochs = 10            # Number of epochs for training
-config.encoder = "resnet18" #resnet34"+config.encoder = "resnet34"
if config.encoder == "resnet18":if config.encoder == "resnet18":
  encoder = models.resnet18     # encoder of unet (contracting path)  encoder = models.resnet18     # encoder of unet (contracting path)
elif config.encoder == "resnet34":elif config.encoder == "resnet34":
  encoder = models.resnet34  encoder = models.resnet34
Expand 13 lines ...
# SWEEPS UNCOMMENT# SWEEPS UNCOMMENT
-config.weight_decay = 0.097 #0.097     # weight decay applied on layers+config.weight_decay = 0.06731 #0.097     # weight decay applied on layers
config.bn_weight_decay = True # whether weight decay is applied on batch norm layersconfig.bn_weight_decay = True # whether weight decay is applied on batch norm layers
config.one_cycle = True         # use the "1cycle" policy -> https://arxiv.org/abs/1803.09820config.one_cycle = True         # use the "1cycle" policy -> https://arxiv.org/abs/1803.09820
# SWEEPS UNCOMMENT# SWEEPS UNCOMMENT
-config.learning_rate = 0.001 #09084 #0.001     # learning rate+config.learning_rate = 0.001367 #0.001     # learning rate
save_model = Falsesave_model = False
# Custom values to filter runs# Custom values to filter runs
# SWEEPS UNCOMMENT# SWEEPS UNCOMMENT
config.training_stages = 2config.training_stages = 2
-src = (SegmentationItemList.from_folder(path_img).use_partial_data(0.5)+src = (SegmentationItemList.from_folder(path_img).use_partial_data(0.2)
Expand 9 lines ...
#src = (SegmentationItemList.from_folder(path_img)#src = (SegmentationItemList.from_folder(path_img)
       .split_by_folder(train='train', valid='val')       .split_by_folder(train='train', valid='val')
       .label_from_func(get_y_fn, classes=segmentation_classes))       .label_from_func(get_y_fn, classes=segmentation_classes))
# Resize, augment, load in batch & normalize (so we can use pre-trained networks)# Resize, augment, load in batch & normalize (so we can use pre-trained networks)
data = (src.transform(get_transforms(), size=config.img_size, tfm_y=True)data = (src.transform(get_transforms(), size=config.img_size, tfm_y=True)
        .databunch(bs=config.batch_size)        .databunch(bs=config.batch_size)
Expand 136 lines ...
Run set2
﻿
﻿
Add a comment