SIDE=224

	def __init__(self,img,grid):

		self.img=img

		self.grid=grid

	def transform(self):

		center=self._getCenter()

		m=getIdentity()

		t1=getTranslation(-center.x,-center.y)

		proj=getProjection()

		rot=getRotation()

		mir=getMirroring()

		for mi in [t1,mir,proj,rot]:

			m=np.matmul(mi,m)

		m=np.matmul(self._computeCrop(m),m)

		img=cv.warpPerspective(self.img,m,(self.SIDE,self.SIDE))

		grid=Corners(c.transform(m) for c in self.grid)

		return (img,list(itertools.chain.from_iterable(grid)))

	def _getCenter(self):

		(a,b,c,d)=self.grid

		p=Line.fromPoints(a,c)

		q=Line.fromPoints(b,d)

		return p.intersect(q)

	def _computeCrop(self,m):

		grid=Corners(c.transform(m) for c in self.grid)

		(x1,y1,x2,y2)=computeBoundingBox(grid)

		(wg,hg)=(x2-x1,y2-y1)

		(left,top,right,bottom)=[random.uniform(0.05,0.2) for i in range(4)]

		t2=getTranslation(left*wg-x1, top*hg-y1)

		scale=getScale(self.SIDE/(wg*(1+left+right)), self.SIDE/(hg*(1+top+bottom)))

		return np.matmul(scale,t2)

	def show(self):

		img=np.copy(self.img)

		for c in self.grid:

			cv.circle(img,(int(c.x),int(c.y)),3,[0,255,0],-1)

		show(img)

def traverseDirs(root):

	stack=[root]

	while len(stack)>0:

		d=stack.pop()

		contents=sorted(os.scandir(d),key=lambda f: f.name,reverse=True)

		if any(f.name=="annotations.json.gz" for f in contents):

			print(d)

			yield d

		for f in contents:

			if f.is_dir(): stack.append(f.path)

def harvestDir(path):

	annotations=DataFile(os.path.join(path,"annotations.json.gz"))

	imgFilter=lambda f: f.is_file() and re.match(r".*\.(jpg|jpeg|png|gif)$", f.name.lower())

	files=sorted(filter(imgFilter,os.scandir(path)),key=lambda f: f.name)

	boards=annotations["."]

	for f in files:

		img=cv.imread(f.path)

		img=cv.cvtColor(img,cv.COLOR_BGR2GRAY)

		for b in boards:

			sample=Sample(img,b.grid)

			(img,label)=sample.transform()

			yield (img,label)

def loadDataset(root):

	testRatio=0.1

	trainRatio=1-testRatio

	images=[]

	labels=[]

	for d in traverseDirs(root):

		for (img,label) in harvestDir(d):

			images.append(img)

			labels.append(label)

	n=len(images)

	keys=list(range(n))

	random.shuffle(keys)

	images=[images[k] for k in keys]

	labels=[labels[k] for k in keys]

	m=int(n*trainRatio)

	return (

		(np.uint8(images[:m]),np.float32(labels[:m])),

		(np.uint8(images[m:]),np.float32(labels[m:]))

	)

def show(img,filename="x"):

	cv.imshow(filename,img)

	cv.waitKey(0)

	cv.destroyAllWindows()

if __name__=="__main__":

import argparse

import numpy as np

from keras.models import load_model

from prepare_data import loadDataset,Sample

from analyzer.epoint import EPoint

from analyzer.corners import Corners

parser=argparse.ArgumentParser()

parser.add_argument("model")

args=parser.parse_args()

model=load_model(args.model)

print("loading data...")

print("done")

for img in testImages:

	label=model.predict(np.reshape(img,(1,224,224,1)))

	print(label)

	points=[]

	for i in range(4):

		points.append(EPoint(label[0][i*2],label[0][i*2+1]))

	corners=Corners(points)

	sample=Sample(np.uint8(img),corners)

	sample.show()

import argparse

from keras.layers import Conv2D,Dropout,Dense,Flatten,MaxPooling2D,BatchNormalization

from keras.models import Sequential,load_model

parser=argparse.ArgumentParser()

parser.add_argument("--load_model")

parser.add_argument("--save_model",default="/tmp/gogo-{0:03}.h5")

parser.add_argument("--epochs",type=int,default=100)

parser.add_argument("--initial_epoch",type=int,default=0)

args=parser.parse_args()

def createFullyConnected():

	model=Sequential([

		Flatten(input_shape=(224,224)),

		Dense(128, activation="relu"),

		Dropout(0.1),

		Dense(64, activation="relu"),

		Dense(8)

	])

	model.compile(

		optimizer='adam',

		loss='mse',

		metrics=['mae','accuracy']

	)

	return model

def createCNN():

	model=Sequential()

	model.add(Conv2D(filters=16,kernel_size=2,padding="same",activation="relu",input_shape=(224,224,1)))

	model.add(Dropout(0.1))

	model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2),padding="valid"))

	model.add(BatchNormalization())

	model.add(Conv2D(32,(5,5),activation="relu"))

	model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2),padding="valid"))

	model.add(Dropout(0.2))

	model.add(BatchNormalization())

	model.add(Conv2D(64,(5,5),activation="relu"))

	model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2),padding="valid"))

	model.add(BatchNormalization())

	model.add(Conv2D(128,(3,3),activation="relu"))

	model.add(MaxPooling2D(pool_size=(2,2),strides=(2,2),padding="valid"))

	model.add(Dropout(0.4))

	model.add(BatchNormalization())

	model.add(Flatten())

	model.add(Dropout(0.1))

	model.add(Dropout(0.1))

	model.add(Dense(8))

	model.compile(optimizer='adam',loss='mse',metrics=['mae','accuracy'])

	return model

model=createCNN()

if args.load_model:

	model=load_model(args.load_model)

for i in range(args.initial_epoch,args.epochs//10):

	model.fit(trainImages.reshape((-1,224,224,1)),trainLabels,epochs=(i+1)*10,initial_epoch=i*10,batch_size=128,validation_split=0.2)