利用Pytorch實(shí)現(xiàn)獲取特征圖的方法詳解
簡(jiǎn)單加載官方預(yù)訓(xùn)練模型
torchvision.models預(yù)定義了很多公開的模型結(jié)構(gòu)
如果pretrained參數(shù)設(shè)置為False,那么僅僅設(shè)定模型結(jié)構(gòu);如果設(shè)置為True,那么會(huì)啟動(dòng)一個(gè)下載流程,下載預(yù)訓(xùn)練參數(shù)
如果只想調(diào)用模型,不想訓(xùn)練,那么設(shè)置model.eval()和model.requires_grad_(False)
想查看模型參數(shù)可以使用modules和named_modules,其中named_modules是一個(gè)長(zhǎng)度為2的tuple,第一個(gè)變量是name,第二個(gè)變量是module本身。
# -*- coding: utf-8 -*-
from torch import nn
from torchvision import models
# load model. If pretrained is True, there will be a downloading process
model = models.vgg19(pretrained=True)
model.eval()
model.requires_grad_(False)
# get model component
features = model.features
modules = features.modules()
named_modules = features.named_modules()
# print modules
for module in modules:
if isinstance(module, nn.Conv2d):
weight = module.weight
bias = module.bias
print(module, weight.shape, bias.shape,
weight.requires_grad, bias.requires_grad)
elif isinstance(module, nn.ReLU):
print(module)
print()
for named_module in named_modules:
name = named_module[0]
module = named_module[1]
if isinstance(module, nn.Conv2d):
weight = module.weight
bias = module.bias
print(name, module, weight.shape, bias.shape,
weight.requires_grad, bias.requires_grad)
elif isinstance(module, nn.ReLU):
print(name, module)
圖片預(yù)處理
使用opencv和pil讀圖都可以使用transforms.ToTensor()把原本[H, W, 3]的數(shù)據(jù)轉(zhuǎn)成[3, H, W]的tensor。但opencv要注意把數(shù)據(jù)改成RGB順序。
vgg系列模型需要做normalization,建議配合torchvision.transforms來(lái)實(shí)現(xiàn)。
mini-batches of 3-channel RGB images of shape (3 x H x W), where H and W are expected to be at least 224. The images have to be loaded in to a range of [0, 1] and then normalized using mean = [0.485, 0.456, 0.406] and std = [0.229, 0.224, 0.225].
參考:https://pytorch.org/hub/pytorch_vision_vgg/
# -*- coding: utf-8 -*-
from PIL import Image
import cv2
import torch
from torchvision import transforms
# transforms for preprocess
preprocess = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
# load image using cv2
image_cv2 = cv2.imread('lena_std.bmp')
image_cv2 = cv2.cvtColor(image_cv2, cv2.COLOR_BGR2RGB)
image_cv2 = preprocess(image_cv2)
# load image using pil
image_pil = Image.open('lena_std.bmp')
image_pil = preprocess(image_pil)
# check whether image_cv2 and image_pil are same
print(torch.all(image_cv2 == image_pil))
print(image_cv2.shape, image_pil.shape)提取單個(gè)特征圖
如果只提取單層特征圖,可以把模型截?cái)?,以?jié)省算力和顯存消耗。
下面索引之所以有+1是因?yàn)閜ytorch預(yù)訓(xùn)練模型里面第一個(gè)索引的module總是完整模塊結(jié)構(gòu),第二個(gè)才開始子模塊。
# -*- coding: utf-8 -*-
from PIL import Image
from torchvision import models
from torchvision import transforms
# load model. If pretrained is True, there will be a downloading process
model = models.vgg19(pretrained=True)
model = model.features[:16 + 1] # 16 = conv3_4
model.eval()
model.requires_grad_(False)
model.to('cuda')
print(model)
# load and preprocess image
preprocess = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
transforms.Resize(size=(224, 224))
])
image = Image.open('lena_std.bmp')
image = preprocess(image)
inputs = image.unsqueeze(0) # add batch dimension
inputs = inputs.cuda()
# forward
output = model(inputs)
print(output.shape)提取多個(gè)特征圖
第一種方式:逐層運(yùn)行model,如果碰到了需要保存的feature map就存下來(lái)。
第二種方式:使用register_forward_hook,使用這種方式需要用一個(gè)類把feature map以成員變量的形式緩存下來(lái)。
兩種方式的運(yùn)行效率差不多
第一種方式簡(jiǎn)單直觀,但是只能處理類似VGG這種沒(méi)有跨層連接的網(wǎng)絡(luò);第二種方式更加通用。
# -*- coding: utf-8 -*-
from PIL import Image
import torch
from torchvision import models
from torchvision import transforms
# load model. If pretrained is True, there will be a downloading process
model = models.vgg19(pretrained=True)
model = model.features[:16 + 1] # 16 = conv3_4
model.eval()
model.requires_grad_(False)
model.to('cuda')
# check module name
for named_module in model.named_modules():
name = named_module[0]
module = named_module[1]
print('-------- %s --------' % name)
print(module)
print()
# load and preprocess image
preprocess = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
transforms.Resize(size=(224, 224))
])
image = Image.open('lena_std.bmp')
image = preprocess(image)
inputs = image.unsqueeze(0) # add batch dimension
inputs = inputs.cuda()
# forward - 1
layers = [2, 7, 8, 9, 16]
layers = sorted(set(layers))
feature_maps = {}
feature = inputs
for i in range(max(layers) + 1):
feature = model[i](feature)
if i in layers:
feature_maps[i] = feature
for key in feature_maps:
print(key, feature_maps.get(key).shape)
# forward - 2
class FeatureHook:
def __init__(self, module):
self.inputs = None
self.output = None
self.hook = module.register_forward_hook(self.get_features)
def get_features(self, module, inputs, output):
self.inputs = inputs
self.output = output
layer_names = ['2', '7', '8', '9', '16']
hook_modules = []
for named_module in model.named_modules():
name = named_module[0]
module = named_module[1]
if name in layer_names:
hook_modules.append(module)
hooks = [FeatureHook(module) for module in hook_modules]
output = model(inputs)
features = [hook.output for hook in hooks]
for feature in features:
print(feature.shape)
# check correctness
for i, layer in enumerate(layers):
feature1 = feature_maps.get(layer)
feature2 = features[i]
print(torch.all(feature1 == feature2))使用第二種方式(register_forward_hook),resnet特征圖也可以順利拿到。
而由于resnet的model已經(jīng)不可以用model[i]的形式索引,所以無(wú)法使用第一種方式。
# -*- coding: utf-8 -*-
from PIL import Image
from torchvision import models
from torchvision import transforms
# load model. If pretrained is True, there will be a downloading process
model = models.resnet18(pretrained=True)
model.eval()
model.requires_grad_(False)
model.to('cuda')
# check module name
for named_module in model.named_modules():
name = named_module[0]
module = named_module[1]
print('-------- %s --------' % name)
print(module)
print()
# load and preprocess image
preprocess = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
transforms.Resize(size=(224, 224))
])
image = Image.open('lena_std.bmp')
image = preprocess(image)
inputs = image.unsqueeze(0) # add batch dimension
inputs = inputs.cuda()
class FeatureHook:
def __init__(self, module):
self.inputs = None
self.output = None
self.hook = module.register_forward_hook(self.get_features)
def get_features(self, module, inputs, output):
self.inputs = inputs
self.output = output
layer_names = [
'conv1',
'layer1.0.relu',
'layer2.0.conv1'
]
hook_modules = []
for named_module in model.named_modules():
name = named_module[0]
module = named_module[1]
if name in layer_names:
hook_modules.append(module)
hooks = [FeatureHook(module) for module in hook_modules]
output = model(inputs)
features = [hook.output for hook in hooks]
for feature in features:
print(feature.shape)問(wèn)題來(lái)了,resnet這種類型的網(wǎng)絡(luò)結(jié)構(gòu)怎么截?cái)啵?/p>
使用如下命令就可以,print查看需要截?cái)嗟侥睦铮缓笥胣n.Sequential重組即可。
需注意重組后網(wǎng)絡(luò)的module_name會(huì)發(fā)生變化。
print(list(model.children()) model = torch.nn.Sequential(*list(model.children())[:6])
以上就是利用Pytorch實(shí)現(xiàn)獲取特征圖的方法詳解的詳細(xì)內(nèi)容,更多關(guān)于Pytorch獲取特征圖的資料請(qǐng)關(guān)注腳本之家其它相關(guān)文章!
相關(guān)文章
Python?torch.onnx.export用法詳細(xì)介紹
這篇文章主要給大家介紹了關(guān)于Python?torch.onnx.export用法詳細(xì)介紹的相關(guān)資料,文中通過(guò)實(shí)例代碼介紹的非常詳細(xì),對(duì)大家的學(xué)習(xí)或者工作具有一定的參考學(xué)習(xí)價(jià)值,需要的朋友可以參考下2022-07-07
python模擬點(diǎn)擊在ios中實(shí)現(xiàn)的實(shí)例講解
在本篇文章里小編給大家整理的是一篇關(guān)于python模擬點(diǎn)擊在ios中實(shí)現(xiàn)的實(shí)例講解內(nèi)容,有需要的朋友們可以參考下。2020-11-11
python使用tkinter包實(shí)現(xiàn)進(jìn)度條
python中的tkinter包是一種常見(jiàn)的設(shè)計(jì)程序的GUI界面用的包,本文將使用tkinter包實(shí)現(xiàn)不同風(fēng)格的進(jìn)度條,感興趣的小伙伴可以跟隨小編一起學(xué)習(xí)一下2024-11-11
python 同時(shí)運(yùn)行多個(gè)程序的實(shí)例
今天小編就為大家分享一篇python 同時(shí)運(yùn)行多個(gè)程序的實(shí)例,具有很好的參考價(jià)值,希望對(duì)大家有所幫助。一起跟隨小編過(guò)來(lái)看看吧2019-01-01

