#!/usr/bin/env python
# coding: utf-8

# In[ ]:


#hide
from utils import *


# # Other computer vision problems

# ## Multi-label classification

# ### The data

# In[ ]:


from fastai2.vision.all import *
path = untar_data(URLs.PASCAL_2007)


# In[ ]:


df = pd.read_csv(path/'train.csv')
df.head()


# ### Sidebar: Pandas and DataFrames

# In[ ]:


df.iloc[:,0]
df.iloc[0,:]
# Trailing ‘:’s are always optional (in numpy, PyTorch, pandas, etc),
#   so this is equivalent:
df.iloc[0]


# In[ ]:


df['fname']


# ### End sidebar

# ### Constructing a data block

# In[ ]:


dblock = DataBlock()


# In[ ]:


dsets = dblock.datasets(df)


# In[ ]:


dsets.train[0]


# In[ ]:


dblock = DataBlock(get_x = lambda r: r['fname'], get_y = lambda r: r['labels'])
dsets = dblock.datasets(df)
dsets.train[0]


# In[ ]:


def get_x(r): return r['fname']
def get_y(r): return r['labels']
dblock = DataBlock(get_x = get_x, get_y = get_y)
dsets = dblock.datasets(df)
dsets.train[0]


# In[ ]:


#hide
Path.BASE_PATH = path


# In[ ]:


def get_x(r): return path/'train'/r['fname']
def get_y(r): return r['labels'].split(' ')
dblock = DataBlock(get_x = get_x, get_y = get_y)
dsets = dblock.datasets(df)
dsets.train[0]


# In[ ]:


dblock = DataBlock(blocks=(ImageBlock, MultiCategoryBlock),
                   get_x = get_x, get_y = get_y)
dsets = dblock.datasets(df)
dsets.train[0]


# In[ ]:


idxs = torch.where(dsets.train[0][1]==1.)[0]
dsets.train.vocab[idxs]


# In[ ]:


def splitter(df):
    train = df.index[~df['is_valid']].tolist()
    valid = df.index[df['is_valid']].tolist()
    return train,valid

dblock = DataBlock(blocks=(ImageBlock, MultiCategoryBlock),
                   splitter=splitter,
                   get_x=get_x, 
                   get_y=get_y)

dsets = dblock.datasets(df)
dsets.train[0]


# In[ ]:


dblock = DataBlock(blocks=(ImageBlock, MultiCategoryBlock),
                   splitter=splitter,
                   get_x=get_x, 
                   get_y=get_y,
                   item_tfms = RandomResizedCrop(128, min_scale=0.35))
dls = dblock.dataloaders(df)


# In[ ]:


dls.show_batch(rows=1, cols=3)


# ### Binary cross entropy

# In[ ]:


learn = cnn_learner(dls, resnet18)


# In[ ]:


x,y = dls.train.one_batch()
activs = learn.model(x)
activs.shape


# In[ ]:


activs[0]


# In[ ]:


def binary_cross_entropy(inputs, targets):
    inputs = inputs.sigmoid()
    return torch.where(targets==1, 1-inputs, inputs).log().mean()


# In[ ]:


loss_func = nn.BCEWithLogitsLoss()
loss = loss_func(activs, y)
loss


# In[ ]:


def say_hello(name, say_what="Hello"): return f"{say_what} {name}."
say_hello('Jeremy'),say_hello('Jeremy', 'Ahoy!')


# In[ ]:


f = partial(say_hello, say_what="Bonjour")
f("Jeremy"),f("Sylvain")


# In[ ]:


learn = cnn_learner(dls, resnet50, metrics=partial(accuracy_multi, thresh=0.2))
learn.fine_tune(3, base_lr=3e-3, freeze_epochs=4)


# In[ ]:


learn.metrics = partial(accuracy_multi, thresh=0.1)
learn.validate()


# In[ ]:


learn.metrics = partial(accuracy_multi, thresh=0.99)
learn.validate()


# In[ ]:


preds,targs = learn.get_preds()


# In[ ]:


accuracy_multi(preds, targs, thresh=0.9, sigmoid=False)


# In[ ]:


xs = torch.linspace(0.05,0.95,29)
accs = [accuracy_multi(preds, targs, thresh=i, sigmoid=False) for i in xs]
plt.plot(xs,accs);


# ## Regression

# ### Assemble the data

# In[ ]:


path = untar_data(URLs.BIWI_HEAD_POSE)


# In[ ]:


#hide
Path.BASE_PATH = path


# In[ ]:


path.ls()


# In[ ]:


(path/'01').ls()


# In[ ]:


img_files = get_image_files(path)
def img2pose(x): return Path(f'{str(x)[:-7]}pose.txt')
img2pose(img_files[0])


# In[ ]:


im = PILImage.create(img_files[0])
im.shape


# In[ ]:


im.to_thumb(160)


# In[ ]:


cal = np.genfromtxt(path/'01'/'rgb.cal', skip_footer=6)
def get_ctr(f):
    ctr = np.genfromtxt(img2pose(f), skip_header=3)
    c1 = ctr[0] * cal[0][0]/ctr[2] + cal[0][2]
    c2 = ctr[1] * cal[1][1]/ctr[2] + cal[1][2]
    return tensor([c1,c2])


# In[ ]:


get_ctr(img_files[0])


# In[ ]:


biwi = DataBlock(blocks=(ImageBlock, PointBlock),
                 get_items=get_image_files,
                 get_y=get_ctr,
                 splitter=FuncSplitter(lambda o: o.parent.name=='13'),
                 batch_tfms=[*aug_transforms(size=(240,320)), Normalize.from_stats(*imagenet_stats)])


# In[ ]:


dls = biwi.dataloaders(path)
dls.show_batch(max_n=9, figsize=(8,6))


# In[ ]:


xb,yb = dls.one_batch()
xb.shape,yb.shape


# In[ ]:


yb[0]


# ### Training a model

# In[ ]:


learn = cnn_learner(dls, resnet18, y_range=(-1,1))


# In[ ]:


def sigmoid_range(x, lo, hi): return torch.sigmoid(x) * (hi-lo) + lo


# In[ ]:


plot_function(partial(sigmoid_range,lo=-1,hi=1), min=-4, max=4)


# In[ ]:


dls.loss_func


# In[ ]:


learn.lr_find()


# In[ ]:


lr = 2e-2
learn.fit_one_cycle(5, lr)


# In[ ]:


math.sqrt(0.0001)


# In[ ]:


learn.show_results(ds_idx=1, max_n=3, figsize=(6,8))


# ## Conclusion

# ## Questionnaire

# ### Further research

# In[ ]: