vgg的grad作为激活值来展示图片物体

Posted 2021-01-14 waldenlake

import torch
import numpy
import torch.nn as nn
import torch.nn.functional as F
from PIL import Image
from torchvision import transforms
import torchvision.models as models

normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],#这是imagenet
                                 std=[0.229, 0.224, 0.225])

tran=transforms.Compose([
    transforms.Resize((224,224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                 std=[0.229, 0.224, 0.225])
])


im=‘./1.jpeg‘
# im=‘./2.jpg‘
im=Image.open(im)
im=tran(im)
im.unsqueeze_(dim=0)

im=torch.autograd.Variable(im,requires_grad=True)


vgg = models.vgg16()
pre=torch.load(‘/home/qk/.torch/models/vgg16-397923af.pth‘)
vgg.load_state_dict(pre)

out=vgg(im)
outnp=out.data[0]
ind=int(numpy.argmax(outnp))
out[0][ind].backward()


grad=im.grad

grad.squeeze_(0)
grad=grad*grad
grad=grad.sum(keepdim=False,dim=0)
grad=torch.sqrt(grad)
rg=torch.max(grad)
grad=grad/rg*255.

#太神奇了，为什么有uint8就能出来激活图，没有就不行！！！！！这要是不搜索一下的话，怎么可能debug出来呢？
# im = Image.fromarray(grad.numpy(),‘L‘) # .eval()  tensor->numpy array

im = Image.fromarray(numpy.uint8(grad.numpy()),‘L‘) # .eval()  tensor->numpy array
im.save(‘grey.png‘)

# input()

from cls import d
print(d[ind])