In [ ]:
%reload_ext autoreload
%autoreload 2
%matplotlib inline
In [ ]:
from fastai.vision import *
from fastai.callbacks.hooks import *
from fastai.utils.mem import *
In [ ]:
path = untar_data(URLs.CAMVID)
path.ls()
Out[ ]:
In [ ]:
path_lbl = path/'labels'
path_img = path/'images'
In [ ]:
# path = Path('./data/camvid-small')
# def get_y_fn(x): return Path(str(x.parent)+'annot')/x.name
# codes = array(['Sky', 'Building', 'Pole', 'Road', 'Sidewalk', 'Tree',
# 'Sign', 'Fence', 'Car', 'Pedestrian', 'Cyclist', 'Void'])
# src = (SegmentationItemList.from_folder(path)
# .split_by_folder(valid='val')
# .label_from_func(get_y_fn, classes=codes))
# bs=8
# data = (src.transform(get_transforms(), tfm_y=True)
# .databunch(bs=bs)
# .normalize(imagenet_stats))
In [ ]:
fnames = get_image_files(path_img)
fnames[:3]
Out[ ]:
In [ ]:
lbl_names = get_image_files(path_lbl)
lbl_names[:3]
Out[ ]:
In [ ]:
img_f = fnames[0]
img = open_image(img_f)
img.show(figsize=(5,5))
In [ ]:
get_y_fn = lambda x: path_lbl/f'{x.stem}_P{x.suffix}'
In [ ]:
mask = open_mask(get_y_fn(img_f))
mask.show(figsize=(5,5), alpha=1)
In [ ]:
src_size = np.array(mask.shape[1:])
src_size,mask.data
Out[ ]:
In [ ]:
codes = np.loadtxt(path/'codes.txt', dtype=str); codes
Out[ ]:
In [ ]:
size = src_size//2
free = gpu_mem_get_free_no_cache()
# the max size of bs depends on the available GPU RAM
if free > 8200: bs=8
else: bs=4
print(f"using bs={bs}, have {free}MB of GPU RAM free")
In [ ]:
src = (SegmentationItemList.from_folder(path_img)
.split_by_fname_file('../valid.txt')
.label_from_func(get_y_fn, classes=codes))
In [ ]:
data = (src.transform(get_transforms(), size=size, tfm_y=True)
.databunch(bs=bs)
.normalize(imagenet_stats))
In [ ]:
data.show_batch(2, figsize=(10,7))
In [ ]:
data.show_batch(2, figsize=(10,7), ds_type=DatasetType.Valid)
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 [ ]:
learn = unet_learner(data, models.resnet34, metrics=metrics, wd=wd)
In [ ]:
lr_find(learn)
learn.recorder.plot()
In [ ]:
lr=3e-3
In [ ]:
learn.fit_one_cycle(10, slice(lr), pct_start=0.9)
In [ ]:
learn.save('stage-1')
In [ ]:
learn.load('stage-1');
In [ ]:
learn.show_results(rows=3, figsize=(8,9))
In [ ]:
learn.unfreeze()
In [ ]:
lrs = slice(lr/400,lr/4)
In [ ]:
learn.fit_one_cycle(12, lrs, pct_start=0.8)
In [ ]:
learn.save('stage-2');
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.
如果内存不够的话，你可能需要重启你的计算内核，然后再返回这一步，同时可能要减少 bs 的设定。
In [ ]:
learn.destroy()
size = src_size
free = gpu_mem_get_free_no_cache()
# the max size of bs depends on the available GPU RAM
if free > 8200: bs=3
else: bs=1
print(f"using bs={bs}, have {free}MB of GPU RAM free")
In [ ]:
data = (src.transform(get_transforms(), size=size, tfm_y=True)
.databunch(bs=bs)
.normalize(imagenet_stats))
In [ ]:
learn = unet_learner(data, models.resnet34, metrics=metrics, wd=wd)
In [ ]:
learn.load('stage-2');
In [ ]:
lr_find(learn)
learn.recorder.plot()
In [ ]:
lr=1e-3
In [ ]:
learn.fit_one_cycle(10, slice(lr), pct_start=0.8)
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)
In [ ]:
learn.save('stage-2-big')
In [ ]:
learn.load('stage-2-big');
In [ ]:
learn.show_results(rows=3, figsize=(10,10))