这是我根据作者你提供代码得到的我的模型结果，难道是我的可视化错了吗？AVT不是一个2分类的问题吗？可视化时 spw ***@***.*** 下面是我的可视化代码： def calculate_metric_percase(pred, gt): pred[pred > 0] = 1 gt[gt > 0] = 1 if pred.sum() > 0 and gt.sum()>0: dice = metric.binary.dc(pred, gt) hd95 = metric.binary.hd95(pred, gt) return dice, hd95 elif pred.sum() > 0 and gt.sum()==0: return 1, 0 else: return 0, 0 import matplotlib.pyplot as plt import matplotlib matplotlib.use("Agg") def visualize_segmentation(image, label, pred, cam_hotmap_list, cam_hotmap_name, save_path, slice_name, model_name): # 定义类别颜色映射 category_colors = { 1: (255, 0, 0), # 红色 (AVT) } # 归一化原始图像到 0-255 范围 image = (image - image.min()) / (image.max() - image.min()) image = (image * 255).astype(np.uint8) # 创建一个彩色图像，初始值是原始灰度图 colored_image = np.stack([image] * 3, axis=-1) # (H, W, 3) colored_image_pre = np.stack([image] * 3, axis=-1) # (H, W, 3) # 替换类别区域的颜色，但背景保留原图 for class_idx, color in category_colors.items(): mask = (label == class_idx) colored_image[mask] = np.array(color, dtype=np.uint8) mask_pre = (pred == class_idx) colored_image_pre[mask_pre] = np.array(color, dtype=np.uint8) # 绘制可视化结果 img_num = 1 fig, axes = plt.subplots(1, 1, figsize=(5 * img_num, 8)) # 右侧显示填充颜色后的图像 axes.imshow(colored_image) # axes[1].set_title(f"{model_name}_Label: ({slice_name})") axes.axis("off") plt.tight_layout() # 保存图片 if save_path is not None: plt.savefig(save_path, dpi=300, bbox_inches='tight') print(f"图片保存成功: {save_path}") #draw slice def draw_slice(net,images, labels, preds, case_name, patch_size=None): net.eval() print(f'images: {images.shape}') print(f'labels: {labels.shape}') print(f'preds: {preds.shape}') print(f'case_name: {case_name}') num, h, w = images.shape for index in range(num): img = images[index] lab = labels[index] if not np.any(lab): continue pred = preds[index] save_path = "./test_draw/14_AVT/" + case_name + "_" + str(index) + ".png" visualize_segmentation(img, lab, pred, [], [], save_path, "slice_" + str(index) + "_" + case_name, model_name="14_AVT") import SimpleITK as sitk from tqdm import tqdm def test_single_volume_AVT(image, label, net, classes, patch_size=[224, 224], test_save_path=None, case=None, origin = None, direction = None, xyz_thickness = None): image, label = image.squeeze(0).cpu().detach().numpy(), label.squeeze(0).cpu().detach().numpy() if len(image.shape) == 3: #3D prediction = np.zeros_like(label) for ind in tqdm(range(image.shape[0])): slice = image[ind, :, :] x, y = slice.shape[0], slice.shape[1] if x != patch_size[0] or y != patch_size[1]: slice = zoom(slice, (patch_size[0] / x, patch_size[1] / y), order=3) # previous using 0 input = torch.from_numpy(slice).unsqueeze(0).unsqueeze(0).float().cuda() net.eval() with torch.no_grad(): outputs = net(input) outputs = outputs[0] out = torch.argmax(torch.softmax(outputs, dim=1), dim=1).squeeze(0) out = out.cpu().detach().numpy() if x != patch_size[0] or y != patch_size[1]: pred = zoom(out, (x / patch_size[0], y / patch_size[1]), order=0) else: pred = out prediction[ind] = pred else: input = torch.from_numpy(image).unsqueeze( 0).unsqueeze(0).float().cuda() net.eval() with torch.no_grad(): out = torch.argmax(torch.softmax(net(input), dim=1), dim=1).squeeze(0) prediction = out.cpu().detach().numpy() draw_slice(net, image, label, prediction, case, patch_size) return 0 # metric_list = [] # for i in range(1, classes): # metric_list.append(calculate_metric_percase(prediction == i, label == i)) # # if test_save_path is not None: # img_itk = sitk.GetImageFromArray(image.astype(np.float32)) # img_itk.SetOrigin(origin) # img_itk.SetDirection(direction) # img_itk.SetSpacing((xyz_thickness[0], xyz_thickness[1], xyz_thickness[2])) # sitk.WriteImage(img_itk, test_save_path + '/' + case + "_img.nii.gz") # # prd_itk = sitk.GetImageFromArray(prediction.astype(np.float32)) # prd_itk.SetOrigin(origin) # prd_itk.SetDirection(direction) # prd_itk.SetSpacing((xyz_thickness[0], xyz_thickness[1], xyz_thickness[2])) # sitk.WriteImage(prd_itk, test_save_path + '/' + case + "_pred.nii.gz") # # lab_itk = sitk.GetImageFromArray(label.astype(np.float32)) # lab_itk.SetOrigin(origin) # lab_itk.SetDirection(direction) # lab_itk.SetSpacing((xyz_thickness[0], xyz_thickness[1], xyz_thickness[2])) # sitk.WriteImage(lab_itk, test_save_path + '/' + case + "_gt.nii.gz") # return metric_list 

…

------------------ 原始邮件 ------------------ 发件人: "HongkunSun/ParaTransCNN" ***@***.***>; 发送时间: 2025年5月21日(星期三) 下午5:16 ***@***.***>; ***@***.******@***.***>; 主题: Re: [HongkunSun/ParaTransCNN] 关于AVT数据集，为什么我可视化出来的数据不是论文中这种血管树？？？？？？ (Issue #13) HongkunSun left a comment (HongkunSun/ParaTransCNN#13) 可视化的是标签还是模型推理出来的结果？数据不会出错的 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

关于AVT数据集，为什么我可视化出来的数据不是论文中这种血管树？？？？？？ #13

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions