飞桨常规赛：PALM眼底彩照视盘探测与分割 - 8月第2名方案

#解压数据!unzip -o data/data86770/seg.zip -d /home/aistudio/work

import randomimport os
random.seed(2021)
mask_dir  = '/home/aistudio/work/seg/Train/masks'img_dir = '/home/aistudio/work/seg/Train/fundus_image'path_list = list()for img in os.listdir(img_dir):
    img_path = os.path.join(img_dir,img)
    mask_path = os.path.join(mask_dir,img.replace('jpg', 'png'))
    path_list.append((img_path, mask_path))
random.shuffle(path_list)
ratio = 0.8train_f = open('/home/aistudio/work/seg/Train/train.txt','w') 
val_f = open('/home/aistudio/work/seg/Train/val.txt' ,'w')for i ,content in enumerate(path_list):
    img, mask = content
    text = img + ' ' + mask + '\n'
    if i < len(path_list) * ratio:
        train_f.write(text)    else:
        val_f.write(text)
train_f.close()
val_f.close()

!pip install paddleseg

#导入常用的库import osimport randomimport numpy as npfrom random import shuffleimport cv2import paddlefrom PIL import Imageimport shutilimport refrom paddle.vision.transforms import functional as Fimport os.path

import paddleseg.transforms as Tfrom paddleseg.datasets import OpticDiscSeg,Dataset

train_transforms = [
    T.RandomHorizontalFlip(),# 水平翻转
    T.RandomVerticalFlip(),# 垂直翻转
    T.RandomDistort(0.4),# 随机扭曲
    T.RandomBlur(0.3),# 高斯模糊
    T.RandomScaleAspect(min_scale=0.5,aspect_ratio=0.5),# 随机缩放
    T.Resize(target_size=(512,512)),
    T.Normalize()  # 图像标准化]
val_transforms = [
    T.Resize(target_size=(512,512)),
    T.Normalize()

]

dataset_root = '/home/aistudio/work/seg/Train'train_path  = '/home/aistudio/work/seg/Train/train.txt'val_path  = '/home/aistudio/work/seg/Train/val.txt'# 构建训练集train_dataset = Dataset(
    dataset_root=dataset_root,
    train_path=train_path,
    transforms=train_transforms,
    num_classes=2,
    mode='train'

                  )#验证集val_dataset = Dataset(
    dataset_root=dataset_root,
    val_path=val_path,
    transforms=val_transforms,
    num_classes=2,
    mode='val'

                  )

import paddlefrom paddleseg.models import UNetfrom paddleseg.models import OCRNetfrom paddleseg.models.losses import CrossEntropyLoss,DiceLoss, MixedLoss

base_lr =0.001iters = 16000 unet_model = UNet(num_classes=2,pretrained='/home/aistudio/output/PLAM_model_3/best_model/model.pdparams')#使用预训练模型unet进行训练# unet_model = UNet(num_classes=2)#使用unet进行训练#自动调整学习率lr =paddle.optimizer.lr.CosineAnnealingDecay(base_lr, T_max=(iters // 3), last_epoch=0.5)
u_optimizer = paddle.optimizer.Adam(lr, parameters=unet_model.parameters())

mixtureLosses = [CrossEntropyLoss(),DiceLoss() ]
mixtureCoef = [0.7,0.3]
losses = {}
losses['types'] = [MixedLoss(mixtureLosses, mixtureCoef)]
losses['coef'] = [1]

#进行训练from paddleseg.core import train
train(
    model = unet_model,
    train_dataset=train_dataset,
    val_dataset=val_dataset,
    optimizer=u_optimizer,
    save_dir='output/PLAM_model_4',
    iters=iters,  
    batch_size=4, 
    save_interval=480,
    log_iters=10,
    num_workers=0,
    losses=losses,
    use_vdl=True
    )

from paddleseg.core import evaluate
model = UNet(num_classes=2)#换自己保存的模型文件model_path = 'output/PLAM_model_4/best_model/model.pdparams'para_state_dict = paddle.load(model_path)
model.set_dict(para_state_dict)
evaluate(model,val_dataset)

%cd ~import randomimport os

test_path = r"work/seg/test"test_lst=[]for test in os.listdir(test_path):              
    test_lst.append(test)  

with open('work/seg/test.txt', 'w') as f:    for line in test_lst:
        f.write(line)
        f.write('\n')

from paddleseg.core import predictimport paddleseg.transforms as pt
transforms = pt.Compose([
    pt.Resize(target_size=(512, 512)),
    pt.Normalize()
])

model = UNet(num_classes=2)#生成图片列表image_list = []with open('work/seg/test.txt' ,'r') as f:    for line in f.readlines():
        image_list.append(os.path.join('work/seg/test/',line.split()[0]))

predict(
        model,        #换自己保存的模型文件
        model_path = 'output/PLAM_model_4/best_model/model.pdparams',
        transforms=transforms,
        image_list=image_list,
        save_dir='results',
    )

!mkdir /home/aistudio/newimport os 
import cv2
result_path = '/home/aistudio/results/pseudo_color_prediction'dist_path = '/home/aistudio/new'for img_name in os.listdir(result_path):
    img_path = os.path.join(result_path, img_name)
    img = cv2.imread(img_path)
    g  = img[:,:,1]
    ret, result = cv2.threshold(g, 127,255, cv2.THRESH_BINARY_INV)
    cv2.imwrite(os.path.join(dist_path,img_name), result)

import cv2import osimport numpy as np'''
图像说明：
图像为二值化图像，255白色为目标物，0黑色为背景
要填充白色目标物中的黑色孔洞
'''def FillHole_1(imgPath,SavePath):
    im_in = cv2.imread(imgPath, cv2.IMREAD_GRAYSCALE);
    th, im_th = cv2.threshold(im_in, 220, 255, cv2.THRESH_BINARY_INV);
    im_floodfill = im_th.copy()
    h, w = im_th.shape[:2]
    mask = np.zeros((h+2, w+2), np.uint8)
    cv2.floodFill(im_floodfill, mask, (0,0), 255);
    im_floodfill_inv = cv2.bitwise_not(im_floodfill)
    im_out = im_th | im_floodfill_inv
    cv2.imwrite(SavePath, im_out)def FillHole_2(imgPath,SavePath):
    im_in = cv2.imread(imgPath, cv2.IMREAD_GRAYSCALE);    # 复制 im_in 图像
    im_floodfill = im_in.copy()    # Mask 用于 floodFill，官方要求长宽+2
    h, w = im_in.shape[:2]
    mask = np.zeros((h+2, w+2), np.uint8)    # floodFill函数中的seedPoint必须是背景
    isbreak = False
    for i in range(im_floodfill.shape[0]):        for j in range(im_floodfill.shape[1]):            if(im_floodfill[i][j]==255):
                seedPoint=(i,j)
                isbreak = True
                break
        if(isbreak):            break
    # 得到im_floodfill
    cv2.floodFill(im_floodfill, mask, seedPoint, 0);    # 得到im_floodfill的逆im_floodfill_inv
    im_floodfill_inv = cv2.bitwise_not(im_floodfill)    # 把im_in、im_floodfill_inv这两幅图像结合起来得到前景
    im_out = im_in | im_floodfill
    cv2.imwrite(SavePath, im_out)


result_path = 'new'dist_path = 'new'for img_name in os.listdir(result_path):
    img_path = os.path.join(result_path, img_name)
    SavePath = os.path.join(dist_path, img_name)

    FillHole_1(img_path,SavePath)
    FillHole_2(img_path,SavePath)

!mkdir /home/aistudio/Disc_Segmentationimport os 
import cv2import matplotlib.pyplot as pltdef cnt_area(cnt):
    area = cv2.contourArea(cnt)    return area

result_path = '/home/aistudio/new'dist_path = '/home/aistudio/Disc_Segmentation'for img_name in os.listdir(result_path):
    img_path = os.path.join(result_path, img_name)
    img = cv2.imread(img_path)
    g  = img[:,:,1]
    ret, threshold = cv2.threshold(g, 127,255, cv2.THRESH_BINARY)


    contours, hierarch = cv2.findContours(threshold, cv2.RETR_LIST, cv2.CHAIN_APPROX_NONE)
    contours.sort(key=cnt_area, reverse=True)    if len(contours) > 1:        for i in range(1,len(contours)):
            cv2.drawContours(threshold, [contours[i]], 0, 0, -1)
    _,result = cv2.threshold(threshold, 127, 255, cv2.THRESH_BINARY_INV)
    cv2.imwrite(os.path.join(dist_path, img_name), result)

# 压缩当前路径所有文件，输出zip文件path='Disc_Segmentation'import zipfile,os
zipName = 'Disc_Segmentation.zip' #压缩后文件的位置及名称f = zipfile.ZipFile( zipName, 'w', zipfile.ZIP_DEFLATED )for dirpath, dirnames, filenames in os.walk(path):    for filename in filenames:        print(filename)
        f.write(os.path.join(dirpath,filename))
f.close()