In [ ]:
import fastai
from fastai.vision import *
from fastai.callbacks import *
from fastai.utils.mem import *
from torchvision.models import vgg16_bn
In [ ]:
path = untar_data(URLs.PETS)
path_hr = path/'images'
path_lr = path/'small-96'
path_mr = path/'small-256'
In [ ]:
il = ImageList.from_folder(path_hr)
In [ ]:
def resize_one(fn, i, path, size):
dest = path/fn.relative_to(path_hr)
dest.parent.mkdir(parents=True, exist_ok=True)
img = PIL.Image.open(fn)
targ_sz = resize_to(img, size, use_min=True)
img = img.resize(targ_sz, resample=PIL.Image.BILINEAR).convert('RGB')
img.save(dest, quality=60)
In [ ]:
# create smaller image sets the first time this nb is run
sets = [(path_lr, 96), (path_mr, 256)]
for p,size in sets:
if not p.exists():
print(f"resizing to {size} into {p}")
parallel(partial(resize_one, path=p, size=size), il.items)
In [ ]:
bs,size=32,128
arch = models.resnet34
src = ImageImageList.from_folder(path_lr).split_by_rand_pct(0.1, seed=42)
In [ ]:
def get_data(bs,size):
data = (src.label_from_func(lambda x: path_hr/x.name)
.transform(get_transforms(max_zoom=2.), size=size, tfm_y=True)
.databunch(bs=bs).normalize(imagenet_stats, do_y=True))
data.c = 3
return data
In [ ]:
data = get_data(bs,size)
In [ ]:
data.show_batch(ds_type=DatasetType.Valid, rows=2, figsize=(9,9))
In [ ]:
t = data.valid_ds[0][1].data
t = torch.stack([t,t])
In [ ]:
def gram_matrix(x):
n,c,h,w = x.size()
x = x.view(n, c, -1)
return (x @ x.transpose(1,2))/(c*h*w)
In [ ]:
gram_matrix(t)
Out[ ]:
In [ ]:
base_loss = F.l1_loss
In [ ]:
vgg_m = vgg16_bn(True).features.cuda().eval()
requires_grad(vgg_m, False)
In [ ]:
blocks = [i-1 for i,o in enumerate(children(vgg_m)) if isinstance(o,nn.MaxPool2d)]
blocks, [vgg_m[i] for i in blocks]
Out[ ]:
In [ ]:
class FeatureLoss(nn.Module):
def __init__(self, m_feat, layer_ids, layer_wgts):
super().__init__()
self.m_feat = m_feat
self.loss_features = [self.m_feat[i] for i in layer_ids]
self.hooks = hook_outputs(self.loss_features, detach=False)
self.wgts = layer_wgts
self.metric_names = ['pixel',] + [f'feat_{i}' for i in range(len(layer_ids))
] + [f'gram_{i}' for i in range(len(layer_ids))]
def make_features(self, x, clone=False):
self.m_feat(x)
return [(o.clone() if clone else o) for o in self.hooks.stored]
def forward(self, input, target):
out_feat = self.make_features(target, clone=True)
in_feat = self.make_features(input)
self.feat_losses = [base_loss(input,target)]
self.feat_losses += [base_loss(f_in, f_out)*w
for f_in, f_out, w in zip(in_feat, out_feat, self.wgts)]
self.feat_losses += [base_loss(gram_matrix(f_in), gram_matrix(f_out))*w**2 * 5e3
for f_in, f_out, w in zip(in_feat, out_feat, self.wgts)]
self.metrics = dict(zip(self.metric_names, self.feat_losses))
return sum(self.feat_losses)
def __del__(self): self.hooks.remove()
In [ ]:
feat_loss = FeatureLoss(vgg_m, blocks[2:5], [5,15,2])
In [ ]:
wd = 1e-3
learn = unet_learner(data, arch, wd=wd, loss_func=feat_loss, callback_fns=LossMetrics,
blur=True, norm_type=NormType.Weight)
gc.collect();
In [ ]:
learn.lr_find()
learn.recorder.plot()
In [ ]:
lr = 1e-3
In [ ]:
def do_fit(save_name, lrs=slice(lr), pct_start=0.9):
learn.fit_one_cycle(10, lrs, pct_start=pct_start)
learn.save(save_name)
learn.show_results(rows=1, imgsize=5)
In [ ]:
do_fit('1a', slice(lr*10))
In [ ]:
learn.unfreeze()
In [ ]:
do_fit('1b', slice(1e-5,lr))
In [ ]:
data = get_data(12,size*2)
In [ ]:
learn.data = data
learn.freeze()
gc.collect()
Out[ ]:
In [ ]:
learn.load('1b');
In [ ]:
do_fit('2a')
In [ ]:
learn.unfreeze()
In [ ]:
do_fit('2b', slice(1e-6,1e-4), pct_start=0.3)
In [ ]:
learn = None
gc.collect();
In [ ]:
256/320*1024
Out[ ]:
In [ ]:
256/320*1600
Out[ ]:
In [ ]:
free = gpu_mem_get_free_no_cache()
# the max size of the test image depends on the available GPU RAM
if free > 8000: size=(1280, 1600) # > 8GB RAM
else: size=( 820, 1024) # <= 8GB RAM
print(f"using size={size}, have {free}MB of GPU RAM free")
In [ ]:
learn = unet_learner(data, arch, loss_func=F.l1_loss, blur=True, norm_type=NormType.Weight)
In [ ]:
data_mr = (ImageImageList.from_folder(path_mr).split_by_rand_pct(0.1, seed=42)
.label_from_func(lambda x: path_hr/x.name)
.transform(get_transforms(), size=size, tfm_y=True)
.databunch(bs=1).normalize(imagenet_stats, do_y=True))
data_mr.c = 3
In [ ]:
learn.load('2b');
In [ ]:
learn.data = data_mr
In [ ]:
fn = data_mr.valid_ds.x.items[0]; fn
Out[ ]:
In [ ]:
img = open_image(fn); img.shape
Out[ ]:
In [ ]:
p,img_hr,b = learn.predict(img)
In [ ]:
show_image(img, figsize=(18,15), interpolation='nearest');
In [ ]:
Image(img_hr).show(figsize=(18,15))