參考:https://github.com/milesial/Pytorch-UNet
實現的是二值汽車圖像語義分割,包括 dense CRF 后處理.
使用python3,我的環境是python3.6
1.使用
1> 預測
1)查看所有的可用選項:
python predict.py -h
返回:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ python predict.py -h usage: predict.py [-h] [--model FILE] --input INPUT [INPUT ...] [--output INPUT [INPUT ...]] [--cpu] [--viz] [--no-save] [--no-crf] [--mask-threshold MASK_THRESHOLD] [--scale SCALE] optional arguments: -h, --help show this help message and exit --model FILE, -m FILE Specify the file in which is stored the model (default : 'MODEL.pth') #指明使用的訓練好的模型文件,默認使用MODEL.pth --input INPUT [INPUT ...], -i INPUT [INPUT ...] #指明要進行預測的圖像文件,必須要有的值 filenames of input images --output INPUT [INPUT ...], -o INPUT [INPUT ...] #指明預測后生成的圖像文件的名字 filenames of ouput images --cpu, -c Do not use the cuda version of the net #指明使用CPU,默認為false,即默認使用GPU --viz, -v Visualize the images as they are processed #當圖像被處理時,將其可視化,默認為false,即不可以可視化 --no-save, -n Do not save the output masks #不存儲得到的預測圖像到某圖像文件中,和--viz結合使用,即可對預測結果可視化,但是不存儲結果,默認為false,即會保存結果 --no-crf, -r Do not use dense CRF postprocessing #指明不使用CRF對輸出進行后處理,默認為false,即使用CRF --mask-threshold MASK_THRESHOLD, -t MASK_THRESHOLD Minimum probability value to consider a mask pixel #最小化考慮掩模像素為白色的概率值,默認為0.5 white --scale SCALE, -s SCALE Scale factor for the input images #輸入圖像的比例因子,默認為0.5
2)預測單一圖片image.jpg並存儲結果到output.jpg的命令
python predict.py -i image.jpg -o output.jpg
測試一下:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ python predict.py --cpu --viz -i image.jpg -o output.jpg Loading model MODEL.pth Using CPU version of the net, this may be very slow Model loaded ! Predicting image image.jpg ... /anaconda3/envs/deeplearning/lib/python3.6/site-packages/torch/nn/modules/upsampling.py:129: UserWarning: nn.Upsample is deprecated. Use nn.functional.interpolate instead. warnings.warn("nn.{} is deprecated. Use nn.functional.interpolate instead.".format(self.name)) /anaconda3/envs/deeplearning/lib/python3.6/site-packages/torch/nn/functional.py:1332: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead. warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.") Visualizing results for image image.jpg, close to continue ...
返回可視化圖片為:
關閉該可視化圖片命令就會運行結束:
Mask saved to output.jpg
(deeplearning) userdeMBP:Pytorch-UNet-master user$
並且在當前文件夾中生成名為output.jpg的文件,該圖為:
3)預測多張圖片並顯示,預測結果不存儲:
python predict.py -i image1.jpg image2.jpg --viz --no-save
測試:
先得到的是image1.jpg的可視化結果:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ python predict.py -i image1.jpg image2.jpg --viz --no-save --cpu Loading model MODEL.pth Using CPU version of the net, this may be very slow Model loaded ! Predicting image image1.jpg ... /anaconda3/envs/deeplearning/lib/python3.6/site-packages/torch/nn/modules/upsampling.py:129: UserWarning: nn.Upsample is deprecated. Use nn.functional.interpolate instead. warnings.warn("nn.{} is deprecated. Use nn.functional.interpolate instead.".format(self.name)) /anaconda3/envs/deeplearning/lib/python3.6/site-packages/torch/nn/functional.py:1332: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead. warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.") Visualizing results for image image1.jpg, close to continue ...
圖為:
關閉這個后就會接着生成image2.jpg的可視化結果:
Predicting image image2.jpg ... Visualizing results for image image2.jpg, close to continue ...
返回圖為:
這時候關閉該可視化服務就會結束了,並且沒有在本地保存生成的圖片
4)如果你的計算機只有CPU,即CPU-only版本,使用選項--cpu指定
5)你可以指定你使用的訓練好的模型文件,使用--mode MODEL.pth
6)如果使用上面的命令選項--no-crf:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ python predict.py -i image1.jpg image2.jpg --viz --no-save --cpu --no-crf
返回的結果是:
還有:
可見crf后處理后,可以將一些不符合事實的判斷結果給剔除,使得結果更加精確
2〉訓練
python train.py -h
首先需要安裝模塊pydensecrf,實現CRF條件隨機場的模塊:
pip install pydensecrf 但是出錯:
pydensecrf/densecrf/include/Eigen/Core:22:10: fatal error: 'complex' file not found #include <complex> ^~~~~~~~~ 1 warning and 1 error generated. error: command 'gcc' failed with exit status 1 ---------------------------------------- Failed building wheel for pydensecrf Running setup.py clean for pydensecrf Failed to build pydensecrf
解決辦法,參考https://github.com/lucasb-eyer/pydensecrf:
先安裝cython,需要0.22以上的版本:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ pip install -U cython Installing collected packages: cython Successfully installed cython-0.29.7
然后從git安裝最新版本:
pip install git+https://github.com/lucasb-eyer/pydensecrf.git
但還是沒有成功
后面找到了新的方法,使用conda來安裝就成功了:
userdeMacBook-Pro:~ user$ conda install -n deeplearning -c conda-forge pydensecrf
-c指明從conda-forge
下載模塊
conda-forge
是可以安裝軟件包的附加渠道,使用該conda-forge
頻道取代defaults
因為直接安裝conda install -n deeplearning pydensecrf找不到該模塊
這時候運行python train.py -h可見支持的選項的信息:
(deeplearning) userdeMBP:Pytorch-UNet-master user$ python train.py -h Usage: train.py [options] Options: -h, --help show this help message and exit -e EPOCHS, --epochs=EPOCHS number of epochs #指明迭代的次數 -b BATCHSIZE, --batch-size=BATCHSIZE batch size #圖像批處理的大小 -l LR, --learning-rate=LR learning rate #使用的學習率 -g, --gpu use cuda #使用GPU進行訓練 -c LOAD, --load=LOAD load file model #下載預訓練的文件,在該基礎上進行訓練 -s SCALE, --scale=SCALE downscaling factor of the images #圖像的縮小因子
3>代碼分析
1》unet定義網絡
unet/unet_parts.py
# sub-parts of the U-Net model import torch import torch.nn as nn import torch.nn.functional as F #實現左邊的橫向卷積 class double_conv(nn.Module): '''(conv => BN => ReLU) * 2''' def __init__(self, in_ch, out_ch): super(double_conv, self).__init__() self.conv = nn.Sequential( #以第一層為例進行講解 #輸入通道數in_ch,輸出通道數out_ch,卷積核設為kernal_size 3*3,padding為1,stride為1,dilation=1 #所以圖中H*W能從572*572 變為 570*570,計算為570 = ((572 + 2*padding - dilation*(kernal_size-1) -1) / stride ) +1 nn.Conv2d(in_ch, out_ch, 3, padding=1), nn.BatchNorm2d(out_ch), #進行批標准化,在訓練時,該層計算每次輸入的均值與方差,並進行移動平均 nn.ReLU(inplace=True), #激活函數 nn.Conv2d(out_ch, out_ch, 3, padding=1), #再進行一次卷積,從570*570變為 568*568 nn.BatchNorm2d(out_ch), nn.ReLU(inplace=True) ) def forward(self, x): x = self.conv(x) return x #實現左邊第一行的卷積 class inconv(nn.Module):# def __init__(self, in_ch, out_ch): super(inconv, self).__init__() self.conv = double_conv(in_ch, out_ch) # 輸入通道數in_ch為3, 輸出通道數out_ch為64 def forward(self, x): x = self.conv(x) return x #實現左邊的向下池化操作,並完成另一層的卷積 class down(nn.Module): def __init__(self, in_ch, out_ch): super(down, self).__init__() self.mpconv = nn.Sequential( nn.MaxPool2d(2), double_conv(in_ch, out_ch) ) def forward(self, x): x = self.mpconv(x) return x #實現右邊的向上的采樣操作,並完成該層相應的卷積操作 class up(nn.Module): def __init__(self, in_ch, out_ch, bilinear=True): super(up, self).__init__() # would be a nice idea if the upsampling could be learned too, # but my machine do not have enough memory to handle all those weights if bilinear:#聲明使用的上采樣方法為bilinear——雙線性插值,默認使用這個值,計算方法為 floor(H*scale_factor),所以由28*28變為56*56 self.up = nn.Upsample(scale_factor=2, mode='bilinear', align_corners=True) else: #否則就使用轉置卷積來實現上采樣,計算式子為 (Height-1)*stride - 2*padding -kernal_size +output_padding self.up = nn.ConvTranspose2d(in_ch//2, in_ch//2, 2, stride=2) self.conv = double_conv(in_ch, out_ch) def forward(self, x1, x2): #x2是左邊特征提取傳來的值 #第一次上采樣返回56*56,但是還沒結束 x1 = self.up(x1) # input is CHW, [0]是batch_size, [1]是通道數,更改了下,與源碼不同 diffY = x1.size()[2] - x2.size()[2] #得到圖像x2與x1的H的差值,56-64=-8 diffX = x1.size()[3] - x2.size()[3] #得到圖像x2與x1的W差值,56-64=-8 #用第一次上采樣為例,即當上采樣后的結果大小與右邊的特征的結果大小不同時,通過填充來使x2的大小與x1相同 #對圖像進行填充(-4,-4,-4,-4),左右上下都縮小4,所以最后使得64*64變為56*56 x2 = F.pad(x2, (diffX // 2, diffX - diffX//2, diffY // 2, diffY - diffY//2)) # for padding issues, see # https://github.com/HaiyongJiang/U-Net-Pytorch-Unstructured-Buggy/commit/0e854509c2cea854e247a9c615f175f76fbb2e3a # https://github.com/xiaopeng-liao/Pytorch-UNet/commit/8ebac70e633bac59fc22bb5195e513d5832fb3bd #將最后上采樣得到的值x1和左邊特征提取的值進行拼接,dim=1即在通道數上進行拼接,由512變為1024 x = torch.cat([x2, x1], dim=1) x = self.conv(x) return x #實現右邊的最高層的最右邊的卷積 class outconv(nn.Module): def __init__(self, in_ch, out_ch): super(outconv, self).__init__() self.conv = nn.Conv2d(in_ch, out_ch, 1) def forward(self, x): x = self.conv(x) return x
unet/unetmodel.py
# full assembly of the sub-parts to form the complete net import torch.nn.functional as F from .unet_parts import * class UNet(nn.Module): def __init__(self, n_channels, n_classes): #圖片的通道數,1為灰度圖像,3為彩色圖像 super(UNet, self).__init__() self.inc = inconv(n_channels, 64) #假設輸入通道數n_channels為3,輸出通道數為64 self.down1 = down(64, 128) self.down2 = down(128, 256) self.down3 = down(256, 512) self.down4 = down(512, 512) self.up1 = up(1024, 256) self.up2 = up(512, 128) self.up3 = up(256, 64) self.up4 = up(128, 64) self.outc = outconv(64, n_classes) def forward(self, x): x1 = self.inc(x) x2 = self.down1(x1) x3 = self.down2(x2) x4 = self.down3(x3) x5 = self.down4(x4) x = self.up1(x5, x4) x = self.up2(x, x3) x = self.up3(x, x2) x = self.up4(x, x1) x = self.outc(x) return F.sigmoid(x) #進行二分類
2》utils
實現dense CRF的代碼utils/crf.py:
詳細可見pydensecrf的使用
#coding:utf-8 import numpy as np import pydensecrf.densecrf as dcrf def dense_crf(img, output_probs): #img為輸入的圖像,output_probs是經過網絡預測后得到的結果 h = output_probs.shape[0] #高度 w = output_probs.shape[1] #寬度 output_probs = np.expand_dims(output_probs, 0) output_probs = np.append(1 - output_probs, output_probs, axis=0) d = dcrf.DenseCRF2D(w, h, 2) #NLABELS=2兩類標注,車和不是車 U = -np.log(output_probs) #得到一元勢 U = U.reshape((2, -1)) #NLABELS=2兩類標注 U = np.ascontiguousarray(U) #返回一個地址連續的數組 img = np.ascontiguousarray(img) d.setUnaryEnergy(U) #設置一元勢 d.addPairwiseGaussian(sxy=20, compat=3) #設置二元勢中高斯情況的值 d.addPairwiseBilateral(sxy=30, srgb=20, rgbim=img, compat=10)#設置二元勢眾雙邊情況的值 Q = d.inference(5) #迭代5次推理 Q = np.argmax(np.array(Q), axis=0).reshape((h, w)) #得列中最大值的索引結果 return Q
utils/utils.py
import random import numpy as np #將圖像分成左右兩塊 def get_square(img, pos): """Extract a left or a right square from ndarray shape : (H, W, C))""" h = img.shape[0] if pos == 0: return img[:, :h] else: return img[:, -h:] def split_img_into_squares(img): return get_square(img, 0), get_square(img, 1) #對圖像進行轉置,將(H, W, C)變為(C, H, W) def hwc_to_chw(img): return np.transpose(img, axes=[2, 0, 1]) def resize_and_crop(pilimg, scale=0.5, final_height=None): w = pilimg.size[0] #得到圖片的寬 h = pilimg.size[1]#得到圖片的高 #默認scale為0.5,即將高和寬都縮小一半 newW = int(w * scale) newH = int(h * scale) #如果沒有指明希望得到的最終高度 if not final_height: diff = 0 else: diff = newH - final_height #重新設定圖片的大小 img = pilimg.resize((newW, newH)) #crop((left,upper,right,lower))函數,從圖像中提取出某個矩形大小的圖像。它接收一個四元素的元組作為參數,各元素為(left, upper, right, lower),坐標系統的原點(0, 0)是左上角 #如果沒有設置final_height,其實就是取整個圖片 #如果設置了final_height,就是取一個上下切掉diff // 2,最后高度為final_height的圖片 img = img.crop((0, diff // 2, newW, newH - diff // 2)) return np.array(img, dtype=np.float32) def batch(iterable, batch_size): """批量處理列表""" b = [] for i, t in enumerate(iterable): b.append(t) if (i + 1) % batch_size == 0: yield b b = [] if len(b) > 0: yield b #然后將數據分為訓練集和驗證集兩份 def split_train_val(dataset, val_percent=0.05): dataset = list(dataset) length = len(dataset) #得到數據集大小 n = int(length * val_percent) #驗證集的數量 random.shuffle(dataset) #將數據打亂 return {'train': dataset[:-n], 'val': dataset[-n:]} #對像素值進行歸一化,由[0,255]變為[0,1] def normalize(x): return x / 255 #將兩個圖片合並起來 def merge_masks(img1, img2, full_w): h = img1.shape[0] new = np.zeros((h, full_w), np.float32) new[:, :full_w // 2 + 1] = img1[:, :full_w // 2 + 1] new[:, full_w // 2 + 1:] = img2[:, -(full_w // 2 - 1):] return new # credits to https://stackoverflow.com/users/6076729/manuel-lagunas def rle_encode(mask_image): pixels = mask_image.flatten() # We avoid issues with '1' at the start or end (at the corners of # the original image) by setting those pixels to '0' explicitly. # We do not expect these to be non-zero for an accurate mask, # so this should not harm the score. pixels[0] = 0 pixels[-1] = 0 runs = np.where(pixels[1:] != pixels[:-1])[0] + 2 runs[1::2] = runs[1::2] - runs[:-1:2] return runs
utils/data_vis.py實現結果的可視化:
import matplotlib.pyplot as plt def plot_img_and_mask(img, mask): fig = plt.figure() a = fig.add_subplot(1, 2, 1) #先是打印輸入的圖片 a.set_title('Input image') plt.imshow(img) b = fig.add_subplot(1, 2, 2) #然后打印預測得到的結果圖片 b.set_title('Output mask') plt.imshow(mask) plt.show()
utils/load.py
# # load.py : utils on generators / lists of ids to transform from strings to # cropped images and masks import os import numpy as np from PIL import Image from .utils import resize_and_crop, get_square, normalize, hwc_to_chw def get_ids(dir): """返回目錄中的id列表""" return (f[:-4] for f in os.listdir(dir)) #圖片名字的后4位為數字,能作為圖片id def split_ids(ids, n=2): """將每個id拆分為n個,為每個id創建n個元組(id, k)""" #等價於for id in ids: # for i in range(n): # (id, i) #得到元祖列表[(id1,0),(id1,1),(id2,0),(id2,1),...,(idn,0),(idn,1)] #這樣的作用是后面會通過后面的0,1作為utils.py中get_square函數的pos參數,pos=0的取左邊的部分,pos=1的取右邊的部分 return ((id, i) for id in ids for i in range(n)) def to_cropped_imgs(ids, dir, suffix, scale): """從元組列表中返回經過剪裁的正確img""" for id, pos in ids: im = resize_and_crop(Image.open(dir + id + suffix), scale=scale) #重新設置圖片大小為原來的scale倍 yield get_square(im, pos) #然后根據pos選擇圖片的左邊或右邊 def get_imgs_and_masks(ids, dir_img, dir_mask, scale): """返回所有組(img, mask)""" imgs = to_cropped_imgs(ids, dir_img, '.jpg', scale) # need to transform from HWC to CHW imgs_switched = map(hwc_to_chw, imgs) #對圖像進行轉置,將(H, W, C)變為(C, H, W) imgs_normalized = map(normalize, imgs_switched) #對像素值進行歸一化,由[0,255]變為[0,1] masks = to_cropped_imgs(ids, dir_mask, '_mask.gif', scale) #對圖像的結果也進行相同的處理 return zip(imgs_normalized, masks) #並將兩個結果打包在一起 def get_full_img_and_mask(id, dir_img, dir_mask): im = Image.open(dir_img + id + '.jpg') mask = Image.open(dir_mask + id + '_mask.gif') return np.array(im), np.array(mask)
3》預測
predict.py使用訓練好的U-net網絡對圖像進行預測,使用dense CRF進行后處理:
#coding:utf-8 import argparse import os import numpy as np import torch import torch.nn.functional as F from PIL import Image from unet import UNet from utils import resize_and_crop, normalize, split_img_into_squares, hwc_to_chw, merge_masks, dense_crf from utils import plot_img_and_mask from torchvision import transforms def predict_img(net, full_img, scale_factor=0.5, out_threshold=0.5, use_dense_crf=True, use_gpu=False): net.eval() #進入網絡的驗證模式,這時網絡已經訓練好了 img_height = full_img.size[1] #得到圖片的高 img_width = full_img.size[0] #得到圖片的寬 img = resize_and_crop(full_img, scale=scale_factor) #在utils文件夾的utils.py中定義的函數,重新定義圖像大小並進行切割,然后將圖像轉為數組np.array img = normalize(img) #對像素值進行歸一化,由[0,255]變為[0,1] left_square, right_square = split_img_into_squares(img)#將圖像分成左右兩塊,來分別進行判斷 left_square = hwc_to_chw(left_square) #對圖像進行轉置,將(H, W, C)變為(C, H, W),便於后面計算 right_square = hwc_to_chw(right_square) X_left = torch.from_numpy(left_square).unsqueeze(0) #將(C, H, W)變為(1, C, H, W),因為網絡中的輸入格式第一個還有一個batch_size的值 X_right = torch.from_numpy(right_square).unsqueeze(0) if use_gpu: X_left = X_left.cuda() X_right = X_right.cuda() with torch.no_grad(): #不計算梯度 output_left = net(X_left) output_right = net(X_right) left_probs = output_left.squeeze(0) right_probs = output_right.squeeze(0) tf = transforms.Compose( [ transforms.ToPILImage(), #重新變成圖片 transforms.Resize(img_height), #恢復原來的大小 transforms.ToTensor() #然后再變成Tensor格式 ] ) left_probs = tf(left_probs.cpu()) right_probs = tf(right_probs.cpu()) left_mask_np = left_probs.squeeze().cpu().numpy() right_mask_np = right_probs.squeeze().cpu().numpy() full_mask = merge_masks(left_mask_np, right_mask_np, img_width)#將左右兩個拆分后的圖片合並起來 #對得到的結果根據設置決定是否進行CRF處理 if use_dense_crf: full_mask = dense_crf(np.array(full_img).astype(np.uint8), full_mask) return full_mask > out_threshold def get_args(): parser = argparse.ArgumentParser() parser.add_argument('--model', '-m', default='MODEL.pth', #指明使用的訓練好的模型文件,默認使用MODEL.pth metavar='FILE', help="Specify the file in which is stored the model" " (default : 'MODEL.pth')") parser.add_argument('--input', '-i', metavar='INPUT', nargs='+', #指明要進行預測的圖像文件 help='filenames of input images', required=True) parser.add_argument('--output', '-o', metavar='INPUT', nargs='+', #指明預測后生成的圖像文件的名字 help='filenames of ouput images') parser.add_argument('--cpu', '-c', action='store_true', #指明使用CPU help="Do not use the cuda version of the net", default=False) parser.add_argument('--viz', '-v', action='store_true', help="Visualize the images as they are processed", #當圖像被處理時,將其可視化 default=False) parser.add_argument('--no-save', '-n', action='store_true', #不存儲得到的預測圖像到某圖像文件中,和--viz結合使用,即可對預測結果可視化,但是不存儲結果 help="Do not save the output masks", default=False) parser.add_argument('--no-crf', '-r', action='store_true', #指明不使用CRF對輸出進行后處理 help="Do not use dense CRF postprocessing", default=False) parser.add_argument('--mask-threshold', '-t', type=float, help="Minimum probability value to consider a mask pixel white", #最小概率值考慮掩模像素為白色 default=0.5) parser.add_argument('--scale', '-s', type=float, help="Scale factor for the input images", #輸入圖像的比例因子 default=0.5) return parser.parse_args() def get_output_filenames(args):#從輸入的選項args值中得到輸出文件名 in_files = args.input out_files = [] if not args.output: #如果在選項中沒有指定輸出的圖片文件的名字,那么就會根據輸入圖片文件名,在其后面添加'_OUT'后綴來作為輸出圖片文件名 for f in in_files: pathsplit = os.path.splitext(f) #將文件名和擴展名分開,pathsplit[0]是文件名,pathsplit[1]是擴展名 out_files.append("{}_OUT{}".format(pathsplit[0], pathsplit[1])) #得到輸出圖片文件名 elif len(in_files) != len(args.output): #如果設置了output名,查看input和output的數量是否相同,即如果input是兩張圖,那么設置的output也必須是兩個,否則報錯 print("Error : Input files and output files are not of the same length") raise SystemExit() else: out_files = args.output return out_files def mask_to_image(mask): return Image.fromarray((mask * 255).astype(np.uint8)) #從數組array轉成Image if __name__ == "__main__": args = get_args() #得到輸入的選項設置的值 in_files = args.input #得到輸入的圖像文件 out_files = get_output_filenames(args) #從輸入的選項args值中得到輸出文件名 net = UNet(n_channels=3, n_classes=1) #定義使用的model為UNet,調用在UNet文件夾下定義的unet_model.py,定義圖像的通道為3,即彩色圖像,判斷類型設為1種 print("Loading model {}".format(args.model)) #指定使用的訓練好的model if not args.cpu: #指明使用GPU print("Using CUDA version of the net, prepare your GPU !") net.cuda() net.load_state_dict(torch.load(args.model)) else: #否則使用CPU net.cpu() net.load_state_dict(torch.load(args.model, map_location='cpu')) print("Using CPU version of the net, this may be very slow") print("Model loaded !") for i, fn in enumerate(in_files): #對圖片進行預測 print("\nPredicting image {} ...".format(fn)) img = Image.open(fn) if img.size[0] < img.size[1]: #(W, H, C) print("Error: image height larger than the width") mask = predict_img(net=net, full_img=img, scale_factor=args.scale, out_threshold=args.mask_threshold, use_dense_crf= not args.no_crf, use_gpu=not args.cpu) if args.viz: #可視化輸入的圖片和生成的預測圖片 print("Visualizing results for image {}, close to continue ...".format(fn)) plot_img_and_mask(img, mask) if not args.no_save:#設置為False,則保存 out_fn = out_files[i] result = mask_to_image(mask) #從數組array轉成Image result.save(out_files[i]) #然后保存 print("Mask saved to {}".format(out_files[i]))
4》訓練
import sys import os from optparse import OptionParser import numpy as np import torch import torch.backends.cudnn as cudnn import torch.nn as nn from torch import optim from eval import eval_net from unet import UNet from utils import get_ids, split_ids, split_train_val, get_imgs_and_masks, batch def train_net(net, epochs=5, batch_size=1, lr=0.1, val_percent=0.05, save_cp=True, gpu=False, img_scale=0.5): dir_img = 'data/train/' #訓練圖像文件夾 dir_mask = 'data/train_masks/' #圖像的結果文件夾 dir_checkpoint = 'checkpoints/' #訓練好的網絡保存文件夾 ids = get_ids(dir_img)#圖片名字的后4位為數字,能作為圖片id #得到元祖列表為[(id1,0),(id1,1),(id2,0),(id2,1),...,(idn,0),(idn,1)] #這樣的作用是后面重新設置生成器時會通過后面的0,1作為utils.py中get_square函數的pos參數,pos=0的取左邊的部分,pos=1的取右邊的部分 #這樣圖片的數量就會變成2倍 ids = split_ids(ids) iddataset = split_train_val(ids, val_percent) #將數據分為訓練集和驗證集兩份 print(''' Starting training: Epochs: {} Batch size: {} Learning rate: {} Training size: {} Validation size: {} Checkpoints: {} CUDA: {} '''.format(epochs, batch_size, lr, len(iddataset['train']), len(iddataset['val']), str(save_cp), str(gpu))) N_train = len(iddataset['train']) #訓練集長度 optimizer = optim.SGD(net.parameters(), #定義優化器 lr=lr, momentum=0.9, weight_decay=0.0005) criterion = nn.BCELoss()#損失函數 for epoch in range(epochs): #開始訓練 print('Starting epoch {}/{}.'.format(epoch + 1, epochs)) net.train() #設置為訓練模式 # reset the generators重新設置生成器 # 對輸入圖片dir_img和結果圖片dir_mask進行相同的圖片處理,即縮小、裁剪、轉置、歸一化后,將兩個結合在一起,返回(imgs_normalized, masks) train = get_imgs_and_masks(iddataset['train'], dir_img, dir_mask, img_scale) val = get_imgs_and_masks(iddataset['val'], dir_img, dir_mask, img_scale) epoch_loss = 0 for i, b in enumerate(batch(train, batch_size)): imgs = np.array([i[0] for i in b]).astype(np.float32) #得到輸入圖像數據 true_masks = np.array([i[1] for i in b]) #得到圖像結果數據 imgs = torch.from_numpy(imgs) true_masks = torch.from_numpy(true_masks) if gpu: imgs = imgs.cuda() true_masks = true_masks.cuda() masks_pred = net(imgs) #圖像輸入的網絡后得到結果masks_pred,結果為灰度圖像 masks_probs_flat = masks_pred.view(-1) #將結果壓扁 true_masks_flat = true_masks.view(-1) loss = criterion(masks_probs_flat, true_masks_flat) #對兩個結果計算損失 epoch_loss += loss.item() print('{0:.4f} --- loss: {1:.6f}'.format(i * batch_size / N_train, loss.item())) optimizer.zero_grad() loss.backward() optimizer.step() print('Epoch finished ! Loss: {}'.format(epoch_loss / i)) #一次迭代后得到的平均損失 if 1: val_dice = eval_net(net, val, gpu) print('Validation Dice Coeff: {}'.format(val_dice)) if save_cp: torch.save(net.state_dict(), dir_checkpoint + 'CP{}.pth'.format(epoch + 1)) print('Checkpoint {} saved !'.format(epoch + 1)) def get_args(): parser = OptionParser() parser.add_option('-e', '--epochs', dest='epochs', default=5, type='int', #設置迭代數 help='number of epochs') parser.add_option('-b', '--batch-size', dest='batchsize', default=10, #設置訓練批處理數 type='int', help='batch size') parser.add_option('-l', '--learning-rate', dest='lr', default=0.1, #設置學習率 type='float', help='learning rate') parser.add_option('-g', '--gpu', action='store_true', dest='gpu', #是否使用GPU,默認是不使用 default=False, help='use cuda') parser.add_option('-c', '--load', dest='load', #下載之前預訓練好的模型 default=False, help='load file model') parser.add_option('-s', '--scale', dest='scale', type='float', #圖像的縮小因子,用來重新設置圖片大小 default=0.5, help='downscaling factor of the images') (options, args) = parser.parse_args() return options if __name__ == '__main__': args = get_args() #得到設置的所有參數信息 net = UNet(n_channels=3, n_classes=1) if args.load: #是否加載預先訓練好的模型 net.load_state_dict(torch.load(args.load)) print('Model loaded from {}'.format(args.load)) if args.gpu: #是否使用GPU,設置為True,則使用 net.cuda() # cudnn.benchmark = True # faster convolutions, but more memory try: #開始訓練 train_net(net=net, epochs=args.epochs, batch_size=args.batchsize, lr=args.lr, gpu=args.gpu, img_scale=args.scale) except KeyboardInterrupt: #如果鍵盤輸入ctrl+c停止,則會將結果保存在INTERRUPTED.pth中 torch.save(net.state_dict(), 'INTERRUPTED.pth') print('Saved interrupt') try: sys.exit(0) except SystemExit: os._exit(0)