Changeset - ecf98a415d97
Parent rev.
Child rev.
[Not reviewed]
default
0 3 0
Laman - 6 years ago 2019-05-09 22:15:27

a different CNN architecture
3 files changed with 40 insertions and 75 deletions:
exp/kerokero/prepare_data.py
8
7
exp/kerokero/test.py
3
4
exp/kerokero/train.py
29
64
0 comments (0 inline, 0 general)
exp/kerokero/prepare_data.py
➞
Show inline comments
 
@@ -20,7 +20,7 @@ class Sample:
 
	SIDE=224
 

			




	
	
	
		
 
	def __init__(self,img,grid):

			




	
	
	
		
 
		""":param img: an image as a 3D np.uint8, channels-last

			




	
	
	
		
 
		""":param img: a greyscale image as a 2D np.uint8

			




	
	
	
		
 
		:param grid: iterable of 4 EPoints, ie. Corners"""

			




	
	
	
		
 
		self.img=img

			




	
	
	
		
 
		self.grid=grid

			




	
	
		
 
@@ -38,7 +38,7 @@ class Sample:

			




	
	
	
		
 
			m=np.matmul(mi,m)

			




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

			




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

			




	
	
	
		
 
		img=cv.cvtColor(img,cv.COLOR_BGR2RGB)

			




	
	
	
		
 
		img=np.float32(img)/128-1

			




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

			




	
	
	
		
 
		grid=list(map(lambda p: 2*p/self.SIDE-EPoint(1,1), grid))

			




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

			




	
	
		
 
@@ -59,7 +59,7 @@ class Sample:

			




	
	
	
		
 
		return np.matmul(scale,t2)

			




	
	
	
		
 

			




	
	
	
		
 
	def show(self):

			




	
	
	
		
 
		img=np.copy(self.img)

			




	
	
	
		
 
		img=cv.cvtColor(self.img,cv.COLOR_GRAY2BGR)

			




	
	
	
		
 
		for c in self.grid:

			




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

			




	
	
	
		
 
		show(img)

			




	
	
		
 
@@ -80,15 +80,16 @@ def traverseDirs(root):

			




	
	
	
		
 
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)[::3]

			




	
	
	
		
 
	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)

			




	
	
	
		
 
			# sample.show()

			




	
	
	
		
 
			(transformedImg,transformedGrid,label)=sample.transform()

			




	
	
	
		
 
			# Sample(transformedImg,map(lambda c: (c+EPoint(1,1))*Sample.SIDE/2,transformedGrid)).show()

			




	
	
	
		
 
			# Sample(np.uint8((transformedImg+1)*128),map(lambda c: (c+EPoint(1,1))*Sample.SIDE/2,transformedGrid)).show()

			




	
	
	
		
 
			yield (transformedImg,label)

			




	
	
	
		
 

			




	
	
	
		
 

			




	
	
		
 
@@ -108,8 +109,8 @@ def loadDataset(root):

			




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

			




	
	
	
		
 
	m=int(n*trainRatio)

			




	
	
	
		
 
	return (

			




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

			




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

			




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

			




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

			




	
	
	
		
 
	)

			




	
	
	
		
 

			




	
	
	
		
 

			




        

    


    
    

    

        

                

                    exp/kerokero/test.py
                

                

                  
                      
                        

                      ➞

                      
                        
                  

                  
                      
                  
                      
                  
                      
                  
                      
                  
                  
                

                

                    Show inline comments
                    
                

        

        

            



	
	
		
 
@@ -3,7 +3,6 @@ import logging as log

			




	
	
	
		
 

			




	
	
	
		
 
import numpy as np

			




	
	
	
		
 
from keras.models import load_model

			




	
	
	
		
 
from keras.applications.inception_v3 import preprocess_input

			




	
	
	
		
 

			




	
	
	
		
 
from prepare_data import loadDataset,Sample

			




	
	
	
		
 
from analyzer.epoint import EPoint

			




	
	
		
 
@@ -25,14 +24,14 @@ with np.load(args.data) as data:

			




	
	
	
		
 
	testLabels=data["testLabels"]

			




	
	
	
		
 
log.info("done")

			




	
	
	
		
 

			




	
	
	
		
 
log.info(model.evaluate(preprocess_input(testImages).reshape((-1,224,224,3)),testLabels))

			




	
	
	
		
 
log.info(model.evaluate(testImages.reshape((-1,224,224,1)),testLabels))

			




	
	
	
		
 

			




	
	
	
		
 
for img in testImages:

			




	
	
	
		
 
	label=model.predict(preprocess_input(np.reshape(img,(1,224,224,3))))

			




	
	
	
		
 
	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]+1)*112,(label[0][i*2+1]+1)*112))

			




	
	
	
		
 
	corners=Corners(points)

			




	
	
	
		
 
	sample=Sample(img,corners)

			




	
	
	
		
 
	sample=Sample(np.uint8((img+1)*128),corners)

			




	
	
	
		
 
	sample.show()

			




        

    


    
    

    

        

                

                    exp/kerokero/train.py
                

                

                  
                      
                        

                      ➞

                      
                        
                  

                  
                      
                  
                      
                  
                      
                  
                      
                  
                  
                

                

                    Show inline comments
                    
                

        

        

            



	
	
		
 
@@ -4,11 +4,9 @@ import argparse

			




	
	
	
		
 
import logging as log

			




	
	
	
		
 

			




	
	
	
		
 
import numpy as np

			




	
	
	
		
 
from keras.layers import Conv2D,Dropout,Dense,Flatten,MaxPooling2D,GlobalAveragePooling2D,BatchNormalization

			




	
	
	
		
 
from keras.models import Sequential,load_model,Model

			




	
	
	
		
 
from keras.optimizers import SGD

			




	
	
	
		
 
from keras.layers import Conv2D,Dropout,Dense,Flatten,MaxPooling2D,BatchNormalization,GlobalAveragePooling2D

			




	
	
	
		
 
from keras.models import Sequential,load_model

			




	
	
	
		
 
from keras.callbacks import TensorBoard

			




	
	
	
		
 
from keras.applications.inception_v3 import InceptionV3,preprocess_input

			




	
	
	
		
 

			




	
	
	
		
 
import config as cfg

			




	
	
	
		
 
import ftp

			




	
	
		
 
@@ -41,84 +39,51 @@ def createFullyConnected():

			




	
	
	
		
 

			




	
	
	
		
 
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(BatchNormalization(input_shape=(224,224,1)))

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Conv2D(24,(5,5),border_mode="same",init="he_normal",activation="relu",input_shape=(224,224,1),dim_ordering="tf"))

			




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

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Conv2D(36,(5,5),activation="relu"))

			




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

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Conv2D(48,(5,5),activation="relu"))

			




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

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Conv2D(64,(3,3),activation="relu"))

			




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

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Conv2D(64,(3,3),activation="relu"))

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(GlobalAveragePooling2D())

			




	
	
	
		
 

			




	
	
	
		
 
	model.add(Dense(500,activation="relu"))

			




	
	
	
		
 
	model.add(Dropout(0.1))

			




	
	
	
		
 
	

			




	
	
	
		
 
	model.add(Dense(128,activation="relu"))

			




	
	
	
		
 
	model.add(Dropout(0.1))

			




	
	
	
		
 
	

			




	
	
	
		
 
	model.add(Dense(90,activation="relu"))

			




	
	
	
		
 
	model.add(Dense(8))

			




	
	
	
		
 

			




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

			




	
	
	
		
 
	model.compile(optimizer="rmsprop",loss="mse",metrics=["mae","accuracy"])

			




	
	
	
		
 
	return model

			




	
	
	
		
 

			




	
	
	
		
 

			




	
	
	
		
 
def createPretrained():

			




	
	
	
		
 
	base=InceptionV3(weights="imagenet",include_top=False,input_shape=(224,224,3))

			




	
	
	
		
 

			




	
	
	
		
 
	x=base.output

			




	
	
	
		
 
	x=GlobalAveragePooling2D()(x)

			




	
	
	
		
 
	x=Dense(1024,activation="relu")(x)

			




	
	
	
		
 
	predictions=Dense(8)(x)

			




	
	
	
		
 

			




	
	
	
		
 
	model=Model(inputs=base.input,outputs=predictions)

			




	
	
	
		
 
	for layer in base.layers:

			




	
	
	
		
 
		layer.trainable=False

			




	
	
	
		
 

			




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

			




	
	
	
		
 
	return model

			




	
	
	
		
 

			




	
	
	
		
 

			




	
	
	
		
 
model=createCNN()

			




	
	
	
		
 
if args.load_model:

			




	
	
	
		
 
	model=load_model(args.load_model)

			




	
	
	
		
 
else:

			




	
	
	
		
 
	model=createPretrained()

			




	
	
	
		
 

			




	
	
	
		
 
log.info("loading data...")

			




	
	
	
		
 
with np.load(args.data) as data:

			




	
	
	
		
 
	trainImages=preprocess_input(data["trainImages"])

			




	
	
	
		
 
	trainImages=data["trainImages"]

			




	
	
	
		
 
	trainLabels=data["trainLabels"]

			




	
	
	
		
 
	testImages=preprocess_input(data["testImages"])

			




	
	
	
		
 
	testImages=data["testImages"]

			




	
	
	
		
 
	testLabels=data["testLabels"]

			




	
	
	
		
 
log.info("done")

			




	
	
	
		
 

			




	
	
	
		
 
tensorboard = TensorBoard(log_dir=os.path.join(args.log_dir,"{}".format(time())))

			




	
	
	
		
 

			




	
	
	
		
 
if not args.load_model:

			




	
	
	
		
 
	model.fit(trainImages.reshape((-1,224,224,3)),trainLabels,epochs=10,batch_size=128,validation_split=0.2,callbacks=[tensorboard])

			




	
	
	
		
 
for layer in model.layers[:249]:

			




	
	
	
		
 
	layer.trainable = False

			




	
	
	
		
 
for layer in model.layers[249:]:

			




	
	
	
		
 
	layer.trainable = True

			




	
	
	
		
 
model.compile(optimizer=SGD(lr=0.0001,momentum=0.9),loss='mse')

			




	
	
	
		
 

			




	
	
	
		
 
tensorboard = TensorBoard(log_dir=os.path.join(cfg.thisDir,"../logs","{}".format(time())))

			




	
	
	
		
 
BIG_STEP=20

			




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

			




	
	
	
		
 
	model.fit(trainImages.reshape((-1,224,224,3)),trainLabels,epochs=(i+1)*BIG_STEP,initial_epoch=i*BIG_STEP,batch_size=128,validation_split=0.2,callbacks=[tensorboard])

			




	
	
	
		
 
	model.fit(trainImages.reshape((-1,224,224,1)),trainLabels,epochs=(i+1)*BIG_STEP,initial_epoch=i*BIG_STEP,batch_size=20,validation_split=0.2,callbacks=[tensorboard])

			




	
	
	
		
 
	path=args.save_model.format((i+1)*BIG_STEP)

			




	
	
	
		
 
	log.info("saving model...")

			




	
	
	
		
 
	model.save(path)

			




	
	
	
		
 
	# ftp.push(path)

			




	
	
	
		
 
log.info(model.evaluate(testImages.reshape((-1,224,224,3)),testLabels))

			




	
	
	
		
 
log.info(model.evaluate(testImages.reshape((-1,224,224,1)),testLabels))

			




        

    



    


    



    



      

      





    
    0 comments (0 inline, 0 general)
    





    


  

  







  


    




    








    
        Server instance: Sirius-538
    
    
        This site is powered by
            Kallithea 0.7.0,
        which is
        © 2010–2023 by various authors & licensed under GPLv3.
            – Support