[个人总结]利用grad-cam实现人民币分类

# -*- coding:utf-8 -*-

import os
import numpy as np
import torch
import cv2
import torch.nn as nn
from torch.utils.data import DataLoader
import torchvision.transforms as transforms
import torchvision.utils as vutils
from torch.utils.tensorboard import SummaryWriter
import torch.optim as optim
from matplotlib import pyplot as plt
import os
from PIL import Image
os.environ ['KMP_DUPLICATE_LIB_OK'] ='True'
import sys
hello_pytorch_DIR = os.path.abspath(os.path.dirname(__file__)+os.path.sep+".."+os.path.sep+"..")
sys.path.append(hello_pytorch_DIR)
fmap_block = list()
grad_block = list()
from model.lenet import LeNet
from tools.my_dataset import RMBDataset
BASE_DIR = os.path.dirname(os.path.abspath(__file__))

torch.manual_seed(1)  # 设置随机种子
rmb_label = {"1": 0, "100": 1}

# 参数设置
MAX_EPOCH = 10
BATCH_SIZE = 16
LR = 0.01
log_interval = 10
val_interval = 1

output_dir = os.path.join(BASE_DIR, "..", "..", "Result", "backward_hook_cam")

fmap_block = list()
input_block = list()
# ============================ step 1/5 数据 ============================

BASE_DIR = os.path.dirname(os.path.abspath(__file__))
split_dir = os.path.abspath(os.path.join(BASE_DIR, "rmb_split"))
if not os.path.exists(split_dir):
    raise Exception(r"数据 {} 不存在, 回到lesson-06\1_split_dataset.py生成数据".format(split_dir))
train_dir = os.path.join(split_dir, "train")
norm_mean = [0.485, 0.456, 0.406]
norm_std = [0.229, 0.224, 0.225]

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, out_dir):
    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)
    path_cam_img = os.path.join(out_dir, "cam1.jpg")
    path_raw_img = os.path.join(out_dir, "raw1.jpg")
    if not os.path.exists(out_dir):
        os.makedirs(out_dir)
    print(cam)
    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, 2).scatter_(1, index, 1)
    one_hot.requires_grad = True
    class_vec = torch.sum(one_hot * outputx)  # 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, (64, 64))
    cam -= np.min(cam)
    cam /= np.max(cam)

    return cam


train_transform = transforms.Compose([
    transforms.Resize((64, 64)),
    transforms.RandomCrop(64, padding=4),
    transforms.RandomGrayscale(p=0.8),
    transforms.ToTensor(),
    transforms.Normalize(norm_mean, norm_std),
])

valid_transform = transforms.Compose([
    transforms.Resize((64, 64)),
    transforms.ToTensor(),
    transforms.Normalize(norm_mean, norm_std),
])

# 构建MyDataset实例
train_data = RMBDataset(data_dir=train_dir, transform=train_transform)

# 构建DataLoder
train_loader = DataLoader(dataset=train_data, batch_size=BATCH_SIZE, shuffle=True)

# ============================ step 2/5 模型 ============================

net = LeNet(classes=2)
net.initialize_weights()

# ============================ step 3/5 损失函数 ============================
criterion = nn.CrossEntropyLoss()                                                   # 选择损失函数

# ============================ step 4/5 优化器 ============================
optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9)                        # 选择优化器
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=10, gamma=0.1)     # 设置学习率下降策略

# ============================ step 5/5 训练 ============================
train_curve = list()

iter_count = 0

for epoch in range(MAX_EPOCH):
    fmap_dict = dict()

    loss_mean = 0.
    correct = 0.
    total = 0.
    net.train()
    for i, data in enumerate(train_loader):
        iter_count += 1
        # forward
        inputs, labels = data
        outputs = net(inputs)
        # backward

        optimizer.zero_grad()
        loss = criterion(outputs, labels)
        loss.backward()

        # update weights
        optimizer.step()
        # 统计分类情况
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).squeeze().sum().numpy()
        # 打印训练信息
        loss_mean += loss.item()
        train_curve.append(loss.item())
        if (i+1) % log_interval == 0:
            loss_mean = loss_mean / log_interval
            print("Training:Epoch[{:0>3}/{:0>3}] Iteration[{:0>3}/{:0>3}] Loss: {:.4f} Acc:{:.2%}".format(
                epoch, MAX_EPOCH, i+1, len(train_loader), loss_mean, correct / total))
            loss_mean = 0.

    scheduler.step()  # 更新学习率
img = cv2.imread('100.jpg', 1)  # H*W*C
x = Image.open('100.jpg').convert('RGB')
norm_mean = [0.485, 0.456, 0.406]
norm_std = [0.229, 0.224, 0.225]
valid_transform = transforms.Compose([
    transforms.Resize((64, 64)),
    transforms.ToTensor(),
    transforms.Normalize(norm_mean, norm_std),
])
x = valid_transform(x)
x.unsqueeze_(0)

net.conv2.register_forward_hook(farward_hook)
net.conv2.register_backward_hook(backward_hook)
outputx = net(x)
net.zero_grad()
class_loss = comp_class_vec(outputx)
class_loss.backward()
grads_val = grad_block[0].cpu().data.numpy().squeeze()
fmap = fmap_block[0].cpu().data.numpy().squeeze()
cam = gen_cam(fmap, grads_val)
img_show = np.float32(cv2.resize(img, (64, 64))) / 255
show_cam_on_image(img_show, cam, output_dir)