下天上的首場比賽開始了首播地的新人啟動(dòng)表。
結(jié)果不咋的。。沒有排名。。
不過不重要，慢慢學(xué)。下面是代碼部分。
1.導(dǎo)入相關(guān)包
import numpy as np
import pandas as pd
import pathlib, sys, os, random, time
import numba, cv2, gc
from tqdm import tqdm_notebook
import matplotlib.pyplot as plt
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
from tqdm.notebook import tqdm
import albumentations as A
import rasterio
from rasterio.windows import Window
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.utils.data as D
import torchvision
from torchvision import transforms as T
重要幾個(gè)的包：
- 數(shù)據(jù)增強(qiáng)：數(shù)據(jù)增強(qiáng)
- torch 相關(guān)：pytorch的內(nèi)核
2.數(shù)據(jù)編碼和解碼
本題學(xué)員需要完成個(gè)人作品表格，需要學(xué)員完成個(gè)人作品表格，并為學(xué)員提供個(gè)人作品，并為學(xué)員提供個(gè)人作品。
編碼后的標(biāo)簽數(shù)據(jù)如下所示：
釋義張圖片的每一行，是第一張圖片的名字，后面跟每張圖片機(jī)構(gòu)的像元號
其編碼和偽裝：
def rle_encode(im):
'''
im: numpy array, 1 - mask, 0 - background
Returns run length as string formated
'''
pixels = im.flatten(order = 'F')
pixels = np.concatenate([[0], pixels, [0]])
runs = np.where(pixels[1:] != pixels[:-1])[0] + 1
runs[1::2] -= runs[::2]
return ' '.join(str(x) for x in runs)
def rle_decode(mask_rle, shape=(512, 512)):
'''
mask_rle: run-length as string formated (start length)
shape: (height,width) of array to return
Returns numpy array, 1 - mask, 0 - background
'''
s = mask_rle.split()
starts, lengths = [np.asarray(x, dtype=int) for x in (s[0:][::2], s[1:][::2])]
starts -= 1
ends = starts + lengths
img = np.zeros(shape[0]*shape[1], dtype=np.uint8)
for lo, hi in zip(starts, ends):
img[lo:hi] = 1
return img.reshape(shape, order='F')
3.定義Dataset類
寫過，數(shù)據(jù)集類是我們讀取數(shù)據(jù)的關(guān)鍵，我們會(huì)在定義中讀取數(shù)據(jù)以及一些類前面的：
class TianChiDataset(D.Dataset):
def __init__(self, paths, rles, transform, test_mode=False):
self.paths = paths
self.rles = rles
self.transform = transform
self.test_mode = test_mode
self.len = len(paths)
self.as_tensor = T.Compose([
T.ToPILImage(),
T.Resize(IMAGE_SIZE),
T.ToTensor(),
T.Normalize([0.625, 0.448, 0.688],
[0.131, 0.177, 0.101]),
])
# get data operation
def __getitem__(self, index):
img = cv2.imread(self.paths[index])
if not self.test_mode:
mask = rle_decode(self.rles[index])
augments = self.transform(image=img, mask=mask)
return self.as_tensor(augments['image']), augments['mask'][None]
else:
return self.as_tensor(img), ''
def __len__(self):
"""
Total number of samples in the dataset
"""
return self.len
上面定義了一個(gè)繼承Pytorch中Dataset的TianChiDataset類。我們先來測試一下。
查看數(shù)據(jù)的編碼和處理是否正確：
train_mask = pd.read_csv('./train_mask.csv', sep='\t', names=['name', 'mask']) # 改路徑
train_mask['name'] = train_mask['name'].apply(lambda x: './train/' + x)
img = cv2.imread(train_mask['name'].iloc[0])
mask = rle_decode(train_mask['mask'].iloc[0])
print(rle_encode(mask) == train_mask['mask'].iloc[0])
如果輸出為真則沒有問題。
定義一些參數(shù)和數(shù)據(jù)增強(qiáng)的方式：
EPOCHES = 20
BATCH_SIZE = 8
IMAGE_SIZE = 256
DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
trfm = A.Compose([
A.Resize(IMAGE_SIZE, IMAGE_SIZE),
A.HorizontalFlip(p=0.5),
A.VerticalFlip(p=0.5),
A.RandomRotate90(),
])
實(shí)例化TianChiDataset類
dataset = TianChiDataset(
train_mask['name'].values,
train_mask['mask'].fillna('').values,
trfm, False
)
以及獲取訓(xùn)練數(shù)據(jù)的第一張圖片標(biāo)簽看下：
image, mask = dataset[0]
plt.figure(figsize=(16,8))
plt.subplot(121)
plt.imshow(mask[0], cmap='gray')
plt.subplot(122)
plt.imshow(image[0]);
4.定義DataLoader
劃分訓(xùn)練集、測試集，并生成DataLoader實(shí)例：
valid_idx, train_idx = [], []
for i in range(len(dataset)):
if i % 7 == 0:
valid_idx.append(i)
# else:
elif i % 7 == 1:
train_idx.append(i)
train_ds = D.Subset(dataset, train_idx)
valid_ds = D.Subset(dataset, valid_idx)
# define training and validation data loaders
loader = D.DataLoader(
train_ds, batch_size=BATCH_SIZE, shuffle=True, num_workers=0)
vloader = D.DataLoader(
valid_ds, batch_size=BATCH_SIZE, shuffle=False, num_workers=0)
5.定義/加載模型
這里我們用的是我們已經(jīng)能夠設(shè)置好的模型，所以最初直接加載torchfcn_resnet50，所以：
def get_model():
model = torchvision.models.segmentation.fcn_resnet50(True)
# pth = torch.load("../input/pretrain-coco-weights-pytorch/fcn_resnet50_coco-1167a1af.pth")
# for key in ["aux_classifier.0.weight", "aux_classifier.1.weight", "aux_classifier.1.bias", "aux_classifier.1.running_mean", "aux_classifier.1.running_var", "aux_classifier.1.num_batches_tracked", "aux_classifier.4.weight", "aux_classifier.4.bias"]:
# del pth[key]
model.classifier[4] = nn.Conv2d(512, 1, kernel_size=(1, 1), stride=(1, 1))
return model
6.損失函數(shù)和優(yōu)化器
@torch.no_grad()
def validation(model, loader, loss_fn):
losses = []
model.eval()
for image, target in loader:
image, target = image.to(DEVICE), target.float().to(DEVICE)
output = model(image)['out']
loss = loss_fn(output, target)
losses.append(loss.item())
return np.array(losses).mean()
model = get_model()
model.to(DEVICE);
optimizer = torch.optim.AdamW(model.parameters(),
lr=1e-4, weight_decay=1e-3)
class SoftDiceLoss(nn.Module):
def __init__(self, smooth=1., dims=(-2,-1)):
super(SoftDiceLoss, self).__init__()
self.smooth = smooth
self.dims = dims
def forward(self, x, y):
tp = (x * y).sum(self.dims)
fp = (x * (1 - y)).sum(self.dims)
fn = ((1 - x) * y).sum(self.dims)
dc = (2 * tp + self.smooth) / (2 * tp + fp + fn + self.smooth)
dc = dc.mean()
return 1 - dc
bce_fn = nn.BCEWithLogitsLoss()
dice_fn = SoftDiceLoss()
def loss_fn(y_pred, y_true):
bce = bce_fn(y_pred, y_true)
dice = dice_fn(y_pred.sigmoid(), y_true)
return 0.8*bce+ 0.2*dice
7.訓(xùn)練并保存最佳模型
到現(xiàn)在我們就可以訓(xùn)練模型了，然后我們最好的了：
header = r'''
Train | Valid
Epoch | Loss | Loss | Time, m
'''
# Epoch metrics time
raw_line = '{:6d}' + '\u2502{:7.3f}'*2 + '\u2502{:6.2f}'
print(header)
EPOCHES = 5
best_loss = 10
for epoch in range(1, EPOCHES+1):
losses = []
start_time = time.time()
model.train()
for image, target in tqdm_notebook(loader):
image, target = image.to(DEVICE), target.float().to(DEVICE)
optimizer.zero_grad()
output = model(image)['out']
loss = loss_fn(output, target)
loss.backward()
optimizer.step()
losses.append(loss.item())
# print(loss.item())
vloss = validation(model, vloader, loss_fn)
print(raw_line.format(epoch, np.array(losses).mean(), vloss,
(time.time()-start_time)/60**1))
losses = []
if vloss < best_loss:
best_loss = vloss
torch.save(model.state_dict(), 'model_best.pth')
8.加載模型及預(yù)測結(jié)果
上面已經(jīng)訓(xùn)練了最好的模型，然后我們只加載模型（也就是驗(yàn)證集的數(shù)據(jù)）就可以了。
加載模型
model.load_state_dict(torch.load("./model_best.pth"))
定義數(shù)據(jù)的操作并進(jìn)行預(yù)測
trfm = T.Compose([
T.ToPILImage(),
T.Resize(IMAGE_SIZE),
T.ToTensor(),
T.Normalize([0.625, 0.448, 0.688],
[0.131, 0.177, 0.101]),
])
subm = []
test_mask = pd.read_csv('./test_a_samplesubmit.csv', sep='\t', names=['name', 'mask'])
test_mask['name'] = test_mask['name'].apply(lambda x: './test_a/' + x)
for idx, name in enumerate(tqdm_notebook(test_mask['name'].iloc[:])):
image = cv2.imread(name)
image = trfm(image)
with torch.no_grad():
image = image.to(DEVICE)[None]
score = model(image)['out'][0][0]
score_sigmoid = score.sigmoid().cpu().numpy()
score_sigmoid = (score_sigmoid > 0.5).astype(np.uint8)
score_sigmoid = cv2.resize(score_sigmoid, (512, 512))
# break
subm.append([name.split('/')[-1], rle_encode(score_sigmoid)])
保存預(yù)測結(jié)果
subm = pd.DataFrame(subm)
subm.to_csv('./tmp.csv', index=None, header=None, sep='\t')
選看一下：
plt.figure(figsize=(16,8))
plt.subplot(121)
plt.imshow(rle_decode(subm[1].fillna('').iloc[0]), cmap='gray')
plt.subplot(122)
plt.imshow(cv2.imread('./test_a/' + subm[0].iloc[0]));
差強(qiáng)人意。。。怪不得沒有效果。。不過我沒有調(diào)參，只是在本機(jī)跑了一個(gè)丹丹。明天優(yōu)化，開始煉制。
參考
【1】官方：aliyunTT6.https://tianchi.com/notebook-ai/detail?spm=517.170ali2.6.67.177c940ali488
【2】網(wǎng)：https://tianchi.yun.com/competition/entrance/531872/信息
