Image segmentation with CamVid¶
In [ ]:
%matplotlib inline
In [ ]:
from fastai2.basics import *
from fastai2.callback.all import *
from fastai2.vision.all import *
from nbdev.showdoc import *
In [ ]:
path = untar_data(URLs.CAMVID)
path.ls()
Out[ ]:
(#4) [Path('/home/sgugger/.fastai/data/camvid/codes.txt'),Path('/home/sgugger/.fastai/data/camvid/labels'),Path('/home/sgugger/.fastai/data/camvid/valid.txt'),Path('/home/sgugger/.fastai/data/camvid/images')]
In [ ]:
path_lbl = path/'labels'
path_img = path/'images'
Data¶
In [ ]:
fnames = get_image_files(path_img)
fnames[:3]
Out[ ]:
(#3) [Path('/home/sgugger/.fastai/data/camvid/images/Seq05VD_f02130.png'),Path('/home/sgugger/.fastai/data/camvid/images/0016E5_05310.png'),Path('/home/sgugger/.fastai/data/camvid/images/0016E5_07230.png')]
In [ ]:
lbl_names = get_image_files(path_lbl)
lbl_names[:3]
Out[ ]:
(#3) [Path('/home/sgugger/.fastai/data/camvid/labels/Seq05VD_f03780_P.png'),Path('/home/sgugger/.fastai/data/camvid/labels/0016E5_00840_P.png'),Path('/home/sgugger/.fastai/data/camvid/labels/0016E5_07997_P.png')]
In [ ]:
img_f = fnames[0]
img = PILImage.create(img_f)
img.show(figsize=(5,5))
Out[ ]:
<matplotlib.axes._subplots.AxesSubplot at 0x7fbc11508950>
In [ ]:
get_y_fn = lambda x: path_lbl/f'{x.stem}_P{x.suffix}'
In [ ]:
mask = PILMask.create(get_y_fn(img_f))
mask.show(figsize=(5,5), alpha=1)
Out[ ]:
<matplotlib.axes._subplots.AxesSubplot at 0x7fbc1143bf10>
In [ ]:
src_size = np.array(mask.shape[1:])
src_size,tensor(mask)
Out[ ]:
(array([960]), tensor([[ 4, 4, 4, ..., 21, 21, 21],
[ 4, 4, 4, ..., 21, 21, 21],
[ 4, 4, 4, ..., 21, 21, 21],
...,
[19, 19, 19, ..., 17, 17, 17],
[30, 19, 19, ..., 17, 17, 17],
[30, 30, 30, ..., 17, 17, 17]], dtype=torch.uint8))
In [ ]:
codes = np.loadtxt(path/'codes.txt', dtype=str); codes
Out[ ]:
array(['Animal', 'Archway', 'Bicyclist', 'Bridge', 'Building', 'Car',
'CartLuggagePram', 'Child', 'Column_Pole', 'Fence', 'LaneMkgsDriv',
'LaneMkgsNonDriv', 'Misc_Text', 'MotorcycleScooter', 'OtherMoving',
'ParkingBlock', 'Pedestrian', 'Road', 'RoadShoulder', 'Sidewalk',
'SignSymbol', 'Sky', 'SUVPickupTruck', 'TrafficCone',
'TrafficLight', 'Train', 'Tree', 'Truck_Bus', 'Tunnel',
'VegetationMisc', 'Void', 'Wall'], dtype='<U17')
Datasets¶
In [ ]:
size = src_size//2
bs=8
In [ ]:
valid_fnames = (path/'valid.txt').read().split('\n')
In [ ]:
# export
def FileSplitter(fname):
"Split `items` depending on the value of `mask`."
valid = Path(fname).read().split('\n')
def _func(x): return x.name in valid
def _inner(o, **kwargs): return FuncSplitter(_func)(o)
return _inner
In [ ]:
camvid = DataBlock(blocks=(ImageBlock, MaskBlock(codes)),
get_items=get_image_files,
splitter=FileSplitter(path/'valid.txt'),
get_y=lambda o: path/'labels'/f'{o.stem}_P{o.suffix}',
batch_tfms=[*aug_transforms(size=(360,480)), Normalize.from_stats(*imagenet_stats)])
In [ ]:
dls = camvid.dataloaders(path/"images", bs=8, path=path)
In [ ]:
show_at(dls.train_ds, 0)
Out[ ]:
<matplotlib.axes._subplots.AxesSubplot at 0x7fbc113dbe90>
In [ ]:
b = dls.train.one_batch()
In [ ]:
b = dls.train.decode(b)
In [ ]:
b[0].shape, b[1].shape
Out[ ]:
(torch.Size([8, 3, 360, 480]), torch.Size([8, 360, 480]))
In [ ]:
dls.show_batch(max_n=4, figsize=(20,14))
Model¶
In [ ]:
name2id = {v:k for k,v in enumerate(codes)}
void_code = name2id['Void']
def acc_camvid(input, target):
target = target.squeeze(1)
mask = target != void_code
return (input.argmax(dim=1)[mask]==target[mask]).float().mean()
In [ ]:
metrics=acc_camvid
# metrics=accuracy
In [ ]:
wd=1e-2
In [ ]:
dls.vocab = codes
In [ ]:
learn = unet_learner(dls, resnet34, metrics=metrics)
In [ ]:
learn.model
Out[ ]:
DynamicUnet(
(layers): ModuleList(
(0): Sequential(
(0): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(4): Sequential(
(0): BasicBlock(
(conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(1): BasicBlock(
(conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(2): BasicBlock(
(conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(5): Sequential(
(0): BasicBlock(
(conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(2): BasicBlock(
(conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(3): BasicBlock(
(conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(6): Sequential(
(0): BasicBlock(
(conv1): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(2): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(3): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(4): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(5): BasicBlock(
(conv1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(7): Sequential(
(0): BasicBlock(
(conv1): Conv2d(256, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(downsample): Sequential(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(1): BasicBlock(
(conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
(2): BasicBlock(
(conv1): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
(3): Sequential(
(0): ConvLayer(
(0): Conv2d(512, 1024, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): ConvLayer(
(0): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
)
(4): UnetBlock(
(shuf): PixelShuffle_ICNR(
(0): ConvLayer(
(0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): PixelShuffle(upscale_factor=2)
)
(bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv1): ConvLayer(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv2): ConvLayer(
(0): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(relu): ReLU()
)
(5): UnetBlock(
(shuf): PixelShuffle_ICNR(
(0): ConvLayer(
(0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): PixelShuffle(upscale_factor=2)
)
(bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv1): ConvLayer(
(0): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(384, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv2): ConvLayer(
(0): Conv2d(384, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(384, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(relu): ReLU()
)
(6): UnetBlock(
(shuf): PixelShuffle_ICNR(
(0): ConvLayer(
(0): Conv2d(384, 768, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(768, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): PixelShuffle(upscale_factor=2)
)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv1): ConvLayer(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv2): ConvLayer(
(0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(relu): ReLU()
)
(7): UnetBlock(
(shuf): PixelShuffle_ICNR(
(0): ConvLayer(
(0): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): PixelShuffle(upscale_factor=2)
)
(bn): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(conv1): ConvLayer(
(0): Conv2d(192, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv2): ConvLayer(
(0): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(96, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(relu): ReLU()
)
(8): PixelShuffle_ICNR(
(0): ConvLayer(
(0): Conv2d(96, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(384, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): PixelShuffle(upscale_factor=2)
)
(9): ResizeToOrig()
(10): MergeLayer()
(11): ResBlock(
(convpath): Sequential(
(0): ConvLayer(
(0): Conv2d(99, 99, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(99, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(1): ConvLayer(
(0): Conv2d(99, 99, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
(1): BatchNorm2d(99, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
(idpath): Sequential()
(act): ReLU(inplace=True)
)
(12): ConvLayer(
(0): Conv2d(99, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)
(1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
)
)
)
In [ ]:
learn.lr_find()
In [ ]:
lr=3e-3
In [ ]:
learn.fit_one_cycle(10, slice(lr), pct_start=0.9, wd=wd)
| epoch | train_loss | valid_loss | acc_camvid | time |
|---|---|---|---|---|
| 0 | 2.228120 | 1.750128 | 0.808884 | 00:51 |
| 1 | 1.873541 | 1.638791 | 0.812702 | 00:38 |
| 2 | 1.619835 | 1.414800 | 0.854545 | 00:38 |
| 3 | 1.432869 | 1.426626 | 0.812345 | 00:39 |
| 4 | 1.237131 | 0.991628 | 0.895642 | 00:39 |
| 5 | 1.040850 | 0.934939 | 0.864443 | 00:39 |
| 6 | 0.866385 | 0.657418 | 0.895164 | 00:39 |
| 7 | 0.719731 | 0.533501 | 0.903053 | 00:39 |
| 8 | 0.617655 | 0.496709 | 0.897753 | 00:40 |
| 9 | 0.510439 | 0.371407 | 0.929200 | 00:40 |
In [ ]:
learn.save('stage-1')
In [ ]:
learn.load('stage-1');
In [ ]:
learn.show_results(max_n=4, figsize=(9,4))
In [ ]:
learn.unfreeze()
In [ ]:
lrs = slice(lr/400,lr/4)
In [ ]:
learn.fit_one_cycle(12, lrs, pct_start=0.8, wd=wd)
| epoch | train_loss | valid_loss | acc_camvid | time |
|---|---|---|---|---|
| 0 | 0.445669 | 0.382609 | 0.925163 | 00:40 |
| 1 | 0.434895 | 0.370302 | 0.928291 | 00:40 |
| 2 | 0.424628 | 0.374690 | 0.923414 | 00:41 |
| 3 | 0.413396 | 0.353807 | 0.929213 | 00:41 |
| 4 | 0.405388 | 0.353545 | 0.927892 | 00:41 |
| 5 | 0.387343 | 0.359148 | 0.924618 | 00:41 |
| 6 | 0.373569 | 0.342784 | 0.924215 | 00:41 |
| 7 | 0.362610 | 0.327022 | 0.928412 | 00:41 |
| 8 | 0.340308 | 0.337757 | 0.923681 | 00:42 |
| 9 | 0.330054 | 0.355201 | 0.917659 | 00:42 |
| 10 | 0.304825 | 0.306595 | 0.928174 | 00:42 |
| 11 | 0.286050 | 0.306130 | 0.928570 | 00:42 |
In [ ]:
learn.save('stage-2');
Go big¶
You may have to restart your kernel and come back to this stage if you run out of memory, and may also need to decrease bs.
In [ ]:
del learn
import gc
gc.collect()
In [ ]:
size = src_size
bs=3
#depending on your GPU RAM you may need to use
# bs=1
print(f"using bs={bs}, have {free}MB of GPU RAM free")
In [ ]:
dls = camvid.dataloaders(path/"images", bs=3, path=path,
batch_tfms=[*aug_transforms(), Normalize.from_stats(*imagenet_stats)])
In [ ]:
dls.vocab = codes
In [ ]:
learn = unet_learner(dls, resnet34, metrics=metrics)
In [ ]:
learn.load('stage-2');
In [ ]:
learn.lr_find()
In [ ]:
lr=1e-3
In [ ]:
learn.fit_one_cycle(10, slice(lr), pct_start=0.8)
| epoch | train_loss | valid_loss | acc_camvid | time |
|---|---|---|---|---|
| 0 | 0.434103 | 0.349006 | 0.919233 | 02:28 |
| 1 | 0.416209 | 0.321492 | 0.926268 | 02:27 |
| 2 | 0.374841 | 0.323334 | 0.928009 | 02:30 |
| 3 | 0.373591 | 0.308711 | 0.925487 | 02:30 |
| 4 | 0.354073 | 0.316937 | 0.921483 | 02:29 |
| 5 | 0.333428 | 0.276603 | 0.931216 | 02:28 |
| 6 | 0.343322 | 0.272028 | 0.929693 | 02:28 |
| 7 | 0.313607 | 0.280513 | 0.929877 | 02:28 |
| 8 | 0.278676 | 0.250869 | 0.932883 | 02:29 |
| 9 | 0.236350 | 0.244575 | 0.935206 | 02:28 |
In [ ]:
learn.save('stage-1-big')
In [ ]:
learn.load('stage-1-big');
In [ ]:
learn.unfreeze()
In [ ]:
lrs = slice(1e-6,lr/10)
In [ ]:
learn.fit_one_cycle(10, lrs)
| epoch | train_loss | valid_loss | acc_camvid | time |
|---|---|---|---|---|
| 0 | 0.223516 | 0.246427 | 0.936520 | 02:30 |
| 1 | 0.237247 | 0.255016 | 0.935370 | 02:39 |
| 2 | 0.225134 | 0.239316 | 0.936160 | 02:44 |
| 3 | 0.222252 | 0.244599 | 0.936024 | 02:43 |
| 4 | 0.216221 | 0.246634 | 0.935405 | 02:42 |
| 5 | 0.223981 | 0.243209 | 0.935413 | 02:42 |
| 6 | 0.206087 | 0.249986 | 0.934875 | 02:42 |
| 7 | 0.201352 | 0.239749 | 0.936708 | 02:41 |
| 8 | 0.201420 | 0.243662 | 0.936199 | 02:41 |
| 9 | 0.201602 | 0.244807 | 0.936131 | 02:41 |
In [ ]:
learn.save('stage-2-big')
In [ ]:
learn.load('stage-2-big');
In [ ]:
learn.show_results(max_n=1, figsize=(20,10), vmin=1, vmax=30)
In [ ]:
fin¶
In [ ]: