[自用代码]pytorch Grad-CAM查看不同模型的激活图

实现了使用grad-cam查看不同网络模型的激活图的功能

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

from __future__ import print_function

import cv2
import os
import numpy as np
from PIL import Image
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision.transforms as transforms
from models import *
import torch.backends.cudnn as cudnn


def img_transform(img_in, transform):
    """
    将img进行预处理，并转换成模型输入所需的形式—— B*C*H*W
    :param img_roi: np.array
    :return:
    """
    img = img_in.copy()
    img = Image.fromarray(np.uint8(img))
    img = transform(img)
    img = img.unsqueeze(0)    # C*H*W --> B*C*H*W
    return img


def img_preprocess(img_in):
    """
    读取图片，转为模型可读的形式
    :param img_in: ndarray, [H, W, C]
    :return: PIL.image
    """
    img = img_in.copy()
    img = cv2.resize(img,(256, 256))
    img = img[:, :, ::-1]   # BGR --> RGB
    transform = transforms.Compose([
        transforms.ToTensor(),
        transforms.Normalize([0.28201964, 0.2816544, 0.21802017],[0.25426927, 0.24146019, 0.1867752])
    ])

    img_input = img_transform(img, transform)
    return img_input


def backward_hook(module, grad_in, grad_out):
    grad_block.append(grad_out[0].detach())


def farward_hook(module, input, output):
    fmap_block.append(output)


def show_cam_on_image(img, mask,ClassName, out_dir,count):
    heatmap = cv2.applyColorMap(np.uint8(255*mask), cv2.COLORMAP_JET)
    heatmap = np.float32(heatmap) / 255
    cam = heatmap + np.float32(img)
    cam = cam / np.max(cam)
    CamName = ClassName + str(count) + "lapcam.jpg"
    RawName = ClassName + str(count) + "raw.jpg"
    path_cam_img = os.path.join(out_dir, CamName)
    path_raw_img = os.path.join(out_dir, RawName)
    if not os.path.exists(out_dir):
        os.makedirs(out_dir)
    cv2.imwrite(path_cam_img, np.uint8(255 * cam))
    cv2.imwrite(path_raw_img, np.uint8(255 * img))


def comp_class_vec(ouput_vec, index=None):
    """
    计算类向量
    :param ouput_vec: tensor
    :param index: int，指定类别
    :return: tensor
    """
    if not index:
        index = np.argmax(ouput_vec.cpu().data.numpy())
    else:
        index = np.array(index)
    index = index[np.newaxis, np.newaxis]
    index = torch.from_numpy(index)
    one_hot = torch.zeros(1, 7).scatter_(1, index, 1)
    one_hot.requires_grad = True
    class_vec = torch.sum(one_hot * output.cpu())  # one_hot = 11.8605

    return class_vec


def gen_cam(feature_map, grads):
    """
    依据梯度和特征图，生成cam
    :param feature_map: np.array， in [C, H, W]
    :param grads: np.array， in [C, H, W]
    :return: np.array, [H, W]
    """
    cam = np.zeros(feature_map.shape[1:], dtype=np.float32)  # cam shape (H, W)

    weights = np.mean(grads, axis=(1, 2))  #

    for i, w in enumerate(weights):
        cam += w * feature_map[i, :, :]

    cam = np.maximum(cam, 0)
    cam = cv2.resize(cam, (256, 256))
    cam -= np.min(cam)
    cam /= np.max(cam)

    return cam

####----------------------------main----------------------------------------####
if __name__ == '__main__':

    
    device = 'cuda' if torch.cuda.is_available() else 'cpu'
    ####----------------------------setting the paths----------------------------####    
    BASE_DIR = os.path.dirname(os.path.abspath(__file__))
    print(BASE_DIR)
    path_img = os.path.join(BASE_DIR, "test_img")
    output_dir = os.path.join(BASE_DIR, "Result", "backward_hook_cam")
    ####---------------------------import classes-------------------------------####
    classes = ('antimonite','arsenopyrite','blende','chalcopyrite','galena','native_gold','pyrite')

    ####--------------------------图片读取；网络加载----------------------------####
    fmap_block = list()
    grad_block = list()

    ####------------------------------Model-------------------------------------####

    # net = vgg16(pretrained=True,NumClass=len(classes))
    # net = resnext50_32x4d(pretrained = True, NumClass=len(classes))
    net = create_RepVGG_B1g2(deploy=False,pretrained = True, clas=len(classes))

    net_name = net.__class__.__name__
    net = net.to(device)
    if device == 'cuda':
        net = torch.nn.DataParallel(net)
        cudnn.benchmark = True
        net.eval()
        
    ####---------------------------Importing weight-----------------------------####
    print('==> Resuming from checkpoint..')
    assert os.path.isdir('checkpoint'), 'Error: no checkpoint directory found!'

    if net_name == 'ResNet':
        path = 'checkpoint\\resnext_acc99.88.pth'
    elif net_name == 'RepVGG':
        path = 'checkpoint\\repvgg_99.82.pth'
    elif net_name == 'VGG':
        path = 'checkpoint\\vgg_acc99.70.pth'
    else:
        print('error')

    checkpoint = torch.load((path), map_location = torch.device('cpu'))
    
    '''
    /home/deyiwang/191112/pytorch-cifar-master/checkpoint/NWPU45
    state = {
        'net': net.state_dict(),
        'acc': acc,
        'epoch': epoch,
    }    
    '''

    net.load_state_dict(checkpoint['net'])

    for root, dirs, files in os.walk(path_img, topdown = True):
        count = 0
        for name in files:
            # 注册hook
            if net_name == 'ResNet':
                net.module.layer4.register_forward_hook(farward_hook)
                net.module.layer4.register_backward_hook(backward_hook)
            elif net_name == 'RepVGG':
                net.module.stage4.register_forward_hook(farward_hook)
                net.module.stage4.register_backward_hook(backward_hook)
            elif net_name == 'VGG':
                net.module.features[40].register_forward_hook(farward_hook)
                net.module.features[40].register_backward_hook(backward_hook)
            else:
                print('error')
            print('======>>>>>>hook done<<<<<<======')
            im_dir = os.path.join(path_img, name)
            ####--------------------------图片读取；网络加载----------------------------####
            fmap_block = list()
            grad_block = list()
            
            img = cv2.imread(str(im_dir), 1)  # H*W*C
            img_input = img_preprocess(img)
            # forward
            output = net(img_input)
            idx = np.argmax(output.cpu().data.numpy())
            print("predict: {}".format(classes[idx]))
        
            # backward
            net.zero_grad()
            class_loss = comp_class_vec(output)
            class_loss.backward()
            print('======>>>>>>backward done<<<<<<======')
            # 生成cam
            grads_val = grad_block[0].cpu().data.numpy().squeeze()
            fmap = fmap_block[0].cpu().data.numpy().squeeze()
            cam = gen_cam(fmap, grads_val)
        
            # 保存cam图片
            ClassName = classes[idx]
            img_show = np.float32(cv2.resize(img, (256, 256))) / 255
            show_cam_on_image(img_show, cam, ClassName, output_dir,count)
            count += 1