太长不看版
1、将PIL.Image转换成批训练的DataLoader
- 为什么一批一批进去处理
2、载入网络( torch.load('Resnet.pkl') ),并将数据放入网络,通过 outputs = model(images) 得到预测值,放在对应的对象中
3、按对象的数字顺序排序,生成热力图保存
耐心看完版
事先准备:
test_path = r'C:\Users\BME419\Desktop\resnet\slide\patch'
background_path = r'E:\WSI\CAMELYON16\Processed\patch-based-classification\raw-data\test\tumor091heatmaps patches\none'
pre_savename = r'C:\Users\BME419\Desktop\resnet\slide\heatmap'
savename = os.path.join(pre_savename, 'heatmap') #保存名字为'heatmap'
batch_size = 64
classes = ['negative','positive']
global positive_prob
positive_prob = [] #positive_prob类型为list
def reload_net(model_name): #可选择四种网络
if model_name == "VGG":
trainednet = torch.load('VGGnet.pkl')
elif model_name == "Google":
trainednet = torch.load('Google.pkl')
elif model_name == "Res":
trainednet =