组合app

egnn_v2/model_type_dict.py

@@ -24,6 +24,7 @@ smov_class_dict = {
         1: 'Mo',
         2: 'V'
 }
+
 norm_line_sv_label = "点线缺陷分类"
 cz_label = "正常掺杂分类"
 xj_label = "012三种不同结构的原子类型分类"
@@ -33,5 +34,26 @@ model_path_dict = {
     cz_label: "model/cz_new.ckpt",
     xj_label: "model/xj_new_2.ckpt",
     smov_label: "model/smov.ckpt"
-
 }
+
+
+from dataclasses import dataclass
+from typing import List, Dict
+
+
+@dataclass
+class ModelLabel:
+    name: str
+    label: List[int]
+
+
+import numpy as np
+
+if __name__ == '__main__':
+    select_point_index = [2, 4]
+    AA =[0, 2, 4]
+    # 创建一个 Person 实例
+    tmp = np.isin(AA, select_point_index)  # 返回 array([False, True, True])
+    atmp = ~np.isin(AA, select_point_index)  # 返回 array([True, False, False])
+    BB = AA[atmp]  # 返回 array([0])
+    print(BB)

egnn_v2/plot_view.py

@@ -10,21 +10,24 @@ from tqdm import tqdm
 from PIL import Image
 from egnn_core.data import load_data
 from egnn_core.data import load_data_v2
+from model_type_dict import ModelLabel
 from sklearn.metrics import accuracy_score, f1_score, recall_score, precision_score, confusion_matrix
+from typing import Dict, List
 
-
+#需要支持不同的size,不止2048
-def plot1(json_path, output_path, psize=24, img_size=2048):
+#img.shape 参数
-    bg = np.zeros((img_size, img_size)) + 255
+def plot1(json_path, output_path, psize=24, img=None):
+    bg = np.zeros(img.shape, np.uint8) + 255
     bg[0, 0] = 0
 
     points, edge_index, labels, _ = load_data(json_path)
 
-    mask_pd = np.zeros((img_size, img_size))
+    mask_pd = np.zeros(img.shape)
     mask_pd[points[:, 0], points[:, 1]] = 1
     mask_pd = np.array(mask_pd, np.uint8)
 
     plt.figure(figsize=(9, 9))
-    plt.imshow(bg, cmap='gray')
+    plt.imshow(img, cmap='gray')
     h, w = np.where(mask_pd != 0)
     plt.scatter(points[:, 1], points[:, 0], s=16, c='#8EB9D9', zorder=2)
 
@@ -139,10 +142,84 @@ def plot_multiple_annotations(img_folder, json_folders, psize=26, output_folder=
             plt.savefig(output_path, dpi=300)
             plt.close()
 
+def plot_arange_combine(img_folder, json_folders: Dict[str, str], select_labels:List[ModelLabel], output_folder: str, psize=40)->None:
+    # 确保输出文件夹存在
+    if not os.path.exists(output_folder):
+        os.makedirs(output_folder, exist_ok=True)
+    c = [
+        ['#FF5733', '#33FF57', '#3357FF'],  # Red, Green, Blue
+        ['#FF33FF', '#33FFFF', '#FFFF33'],  # Magenta, Cyan, Yellow
+        ['#F08080', '#90EE90', '#ADD8E6'],  # Light Coral, Light Green, Light Blue
+        ['#FFD700', '#8A2BE2', '#FF4500']  # Gold, Blue Violet, Orange Red
+    ]
 
+    # 遍历指定文件夹中的所有文件
+    for img_filename in os.listdir(img_folder):
+        plt.figure(figsize=(9, 9))
+
+        if img_filename.endswith(".jpg"):  # 确保是jpeg文件
+            img_path = os.path.join(img_folder, img_filename)
+            img = cv2.imread(img_path, 0)
+            base_points_indexs = []
+            for i, select_label in enumerate(select_labels):
+                # 获取json文件
+                json_path = os.path.join(json_folders[select_label.name], os.path.splitext(os.path.basename(img_filename))[0] + '.json')
+                points, edge_index, labels, _ = load_data_v2(json_path)
+                # 第一个要特殊处理，要把全部标签点都画出来
+                if i == 0:
+                    # 获取基础标签，可能有多个，所以用数组
+
+                    for j, label in enumerate(select_label.label):
+                        select_point_index = np.where(labels == select_label.label[j])[0]
+                        base_points_indexs.append(select_point_index)
+
+                    mask_pd = np.zeros(img.shape, np.uint8)
+                    mask_pd[points[:, 0], points[:, 1]] = labels + 1
+                    mask_pd = np.array(mask_pd, np.uint8)
+
+                    for q in range(3):
+                        h, w = np.where(mask_pd == q + 1)
+                        plt.scatter(w, h, s=psize, c=c[i][q])
+                else:
+                    model_label_index = []
+                    # 获取用户选择的标签对应的点
+                    for k, sub_select_label in enumerate(select_label.label):
+                        tmp = np.where(labels == sub_select_label)[0]
+                        model_label_index.append(tmp)
+
+                    # 计算补集，去除基础标签的点
+                    for index in model_label_index:
+                        indexs = []
+                        for base_points_index in base_points_indexs:
+                            index = index[~np.isin(index, base_points_index)]
+                            indexs.append(index)
+
+
+                    for index in indexs:
+                        labelsv = labels[index]
+                        pointsv = points[index]
+                        mask_pd = np.zeros(img.shape, np.uint8)
+                        mask_pd[pointsv[:, 0], pointsv[:, 1]] = labelsv + 1
+                        mask_pd = np.array(mask_pd, np.uint8)
+                        for q in range(3):
+                            h, w = np.where(mask_pd == q + 1)
+                            plt.scatter(w, h, s=psize, c=c[i][q])
+
+            plt.imshow(img, cmap='gray')
+            plt.axis('off')
+            plt.tight_layout()
+            plt.savefig('kw.jpg', dpi=300)
+            plt.show()
+
+
+
+
+
+#需求：根据用户筛选的标签，画出对应的图像
 def plot_combine(img_path, json_path, json_path2, json_path3, json_path4, psize=40):
     c = ['#dfc8a0', '#814d81', '#debd97']
     ##814d81紫色
+    #根据预测结果获取标签
     points, edge_index, labels, _ = load_data_v2(json_path)
     img = cv2.imread(img_path, 0)
     select_point_index = np.where(labels == 1)[0]
@@ -291,9 +368,30 @@ if __name__ == '__main__':
             '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/012三种不同结构的原子类型分类/gnn_predict_post_process/49.json',
         ''
         ]
+        new_json_folder = [
+            '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/012三种不同结构的原子类型分类/gnn_predict_post_process/49.json',
+            '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/SMoV三种原子分类/gnn_predict_post_process/49.json',
+            '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/正常掺杂分类/gnn_predict_post_process/49.json',
+            ''
+        ]
+
+        dict_json_folder = {
+            "正常掺杂分类": '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/正常掺杂分类/gnn_predict_post_process',
+            "SMoV三种原子分类": '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/SMoV三种原子分类/gnn_predict_post_process',
+            "012三种不同结构的原子类型分类": '/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/012三种不同结构的原子类型分类/gnn_predict_post_process',
+        }
+
+        select_labels = [
+            ModelLabel(name="012三种不同结构的原子类型分类", label=[1]),
+            ModelLabel(name="SMoV三种原子分类", label=[0,1]),
+            ModelLabel(name="正常掺杂分类", label=[1]),
+        ]
+
+        plot_arange_combine(folder, dict_json_folder, select_labels=select_labels, output_folder='/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/combine_view_0731')
         #plot2(folder, folder)
-        plot_combine(img_folder, json_folders[0], json_folders[1], json_folders[2], json_folders[3])
+        #plot_combine(img_folder, new_json_folder[0], new_json_folder[1], new_json_folder[2], new_json_folder[3])
         #plot_multiple_annotations(img_folder, json_folders, output_folder='/home/deploy/script/bohrim-app/bohrium-app/pythonProject/msunet/result_combine_3/combine_view')
+
     except Exception as e:
         import traceback
         print(f"Error processing error: {e}")

egnn_v2/plot_vor_xj_02(1).py

@@ -0,0 +1,108 @@
+import os
+import cv2
+import glob
+import json
+import numpy as np
+import pandas as pd
+import matplotlib.pyplot as plt
+
+from tqdm import tqdm
+from PIL import Image
+from core.data import load_data_v2_plot,load_data_v2_plot_jj
+from sklearn.metrics import accuracy_score, f1_score, recall_score, precision_score, confusion_matrix
+
+
+def plot1(json_path, psize=24, img_size=2048):
+    bg = np.zeros((img_size, img_size)) + 255
+    bg[0, 0] = 0
+
+    points, edge_index, labels, _ = load_data_v2_plot(json_path,max_edge_length=28)
+
+    mask_pd = np.zeros((img_size, img_size))
+    mask_pd[points[:, 0], points[:, 1]] = 1
+    mask_pd = np.array(mask_pd, np.uint8)
+
+    plt.figure(figsize=(9, 9))
+    plt.imshow(bg, cmap='gray')
+    h, w = np.where(mask_pd != 0)
+    plt.scatter(points[:, 1], points[:, 0], s=16, c='#8EB9D9', zorder=2)
+    #
+    for idx, (s, e) in enumerate(edge_index.T):
+        s = points[s]
+        e = points[e]
+        plt.plot([s[1], e[1]], [s[0], e[0]], linewidth=1, c='#C0C0C0', zorder=1)
+
+    plt.axis('off')
+
+    plt.tight_layout()
+
+
+def plot2(img_path, json_path, psize=40):
+    # c = ['#84aad0', '#84aad0', '#84aad0']
+    c = [ '#dfc8a0', '#84aad0','#93ba66']
+
+    # c = [ '#dfc8a0','#debd97', '#84aad0']
+    # c = ['#9BB6CF', '#9BB6CF', '#9BB6CF']
+    # c = ['#83a2c1', '#9BB6CF', '#9BB6CF']
+
+
+    img = cv2.imread(img_path, 0)
+
+    # points, edge_index, labels, _ = load_data_v2_plot(json_path,max_edge_length=25)
+    points, edge_index, labels, _ = load_data_v2_plot(json_path,max_edge_length=28)
+    # points, edge_index, labels, _ = load_data_v2_plot_jj(json_path,max_edge_length=31)
+
+    mask_pd = np.zeros(img.shape)
+
+    # mask_pd = np.zeros((512, 512)) + 256
+    # mask_pd = np.zeros((w, h)) + 256
+    mask_pd[points[:, 0], points[:, 1]] = labels + 1
+    mask_pd = np.array(mask_pd, np.uint8)
+
+    plt.figure(figsize=(9, 9))
+    plt.imshow(img, cmap='gray')
+
+    # img = np.array(Image.open(img_path))
+    # bg = np.zeros_like(img) + 255
+    # bg[0, 0] = 0
+    # plt.imshow(bg, cmap='gray')
+
+    for idx, (s, e) in enumerate(edge_index.T):
+        s = points[s]
+        e = points[e]
+        plt.plot([s[1], e[1]], [s[0], e[0]], linewidth=1, c='#C0C0C0', zorder=1)
+
+    for i in range(3):
+        h, w = np.where(mask_pd == i + 1)
+        plt.scatter(w, h, s=psize, c=c[i])
+
+
+    plt.axis('off')
+
+    plt.tight_layout()
+    plt.savefig('2.jpg',dpi=300)
+    plt.show()
+
+# json_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/xj_gnn/train/raw/train1.json'
+# img_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/xj_gnn/train/raw/train1.jpg'
+# json_path = '/home/gao/mouclear/cc/data/e2e-xj-adjust/xj-adjust/atom/test-5.json'
+# img_path = '/home/gao/mouclear/cc/data/e2e-xj-adjust/xj-adjust/atom/test-5.jpg'
+# json_path = '/home/gao/mouclear/cc/data/e2e_xj_final/xj_result/raw/test.json'
+# img_path = '/home/gao/mouclear/cc/data/e2e_xj_final/xj_result/raw/test.jpg'
+# json_path = '/home/gao/mouclear/cc/data/e2e_xj_final/atom/test.json'
+# img_path = '/home/gao/mouclear/cc/data/e2e_xj_final/atom/test.jpg'a,m,,,,,
+# json_path = '/home/gao/mouclear/cc/code/pd_2/data/gnn_vor/train/raw/1.json'
+# img_path = '/home/gao/mouclear/cc/code/pd_2/data/gnn_vor/train/raw/1.jpg'
+json_path = '/home/gao/mouclear/cc/final_todo/fig_50/53-1/xj/after_norm/53-1.json'
+img_path = '/home/gao/mouclear/cc/final_todo/fig_50/53-1/xj/after_norm/53-1.jpg'
+# json_path = '/home/gao/mouclear/cc/data/e2e-xj-adjust/xj-adjust/atom/another/test.json'
+# img_path = '/home/gao/mouclear/cc/data/e2e-xj-adjust/xj-adjust/atom/another/test.jpg'
+# json_path = '/home/gao/mouclear/cc/data/edge/point-2.json'
+# img_path = '/home/gao/mouclear/cc/data/edge/point-2.jpg'
+# json_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/end-to-end-gnn/raw/test.json'
+# img_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/end-to-end-gnn/raw/test.jpg'
+# # json_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/end-to-end-gnn/gt/test.json'
+# img_path = '/home/gao/mouclear/cc/data/end-to-end-result-xj/end-to-end-gnn/gt/test.jpg'
+
+plot2(img_path, json_path)
+# plot(img_path, json_path)

msunet/app.py

@@ -81,8 +81,8 @@ class CombineGnnPredictOptions(BaseModel):
     select_models: List[SingleWrapper] = Field(default=[SingleWrapper(model_name=norm_line_sv_label)],min_items=1, unique_items=True, description="选择的模型")
     edges_length: Float = Field(default=35.5, description="连接节点边长")
     data_path: InputFilePath = Field(..., ftypes=['zip'], description="测试的数据集")
-    model_path: Optional[InputFilePath] = Field(ftypes=['ckpt'],  description="原子缺陷检测模型")
+    #model_path: Optional[InputFilePath] = Field(ftypes=['ckpt'],  description="原子缺陷检测模型")
-    gnn_model_path: Optional[InputFilePath] = Field(ftypes=['ckpt'], description="gnn模型")
+    #gnn_model_path: Optional[InputFilePath] = Field(ftypes=['ckpt'], description="gnn模型")
     output_dir: OutputDirectory = Field(
         default="./result"
     )  # default will be override after online