hsv_1

Former-commit-id: 57a8fb7799ea7543b4af1ea49626346d50546f82

Picture_Train/main.py

@@ -1,16 +1,68 @@
-# 这是一个示例 Python 脚本。
+import cv2
+import numpy as np
+from pathlib import Path
+from train_2 import train_model,prepare_data,predict
+import joblib
 
-# 按 Shift+F10 执行或将其替换为您的代码。
-# 按 双击 Shift 在所有地方搜索类、文件、工具窗口、操作和设置。
+inp = Path("data/train")
+val = Path("data/val")
+proc = Path("data/proc")
+proc.mkdir(exist_ok=True)
 
 
-def print_hi(name):
-    # 在下面的代码行中使用断点来调试脚本。
-    print(f'Hi, {name}')  # 按 Ctrl+F8 切换断点。
+def preproc(dir:Path):
+    d={}
+    for file in dir.glob("*/*.jpg"):
+        im = cv2.imread(file)
+        if im is None:
+            print(f"Error reading image: {file}")
+            continue
+        cl = file.parents[0].name
+        if cl not in d:
+            d[cl] = []
+        hsv = cv2.cvtColor(im,cv2.COLOR_BGR2HSV)
+        mask = hsv[:,:,1] > 150
+        # cor = np.argwhere(mask)
+        # y_min,x_min = cor.min(axis=0)
+        # y_max,x_max = cor.max(axis=0)
+        mask = mask[:,:,np.newaxis]
+        # cv2.findCoun
+        # im = im[y_min:y_max,x_min:x_max]
+        cnt = np.count_nonzero(mask)
+        hsv*=mask
+        h = round(np.sum(hsv[:,:,0])/cnt)
+        s = round(np.sum(hsv[:,:,1])/cnt)
+        v = round(np.sum(hsv[:,:,2])/cnt)
+        name = f"{h}_{s}_{v}.jpg"
+        d[cl].append((h,s,v))
+        # (proc/cl).mkdir(exist_ok=True)
+        # cv2.imwrite(proc/cl/name,cv2.cvtColor(hsv,cv2.COLOR_HSV2BGR))
+    return d
 
+d:dict[str,list[tuple[int,int,int]]] = preproc(inp)
+val:dict[str,list[tuple[int,int,int]]] = preproc(val)
 
-# 按装订区域中的绿色按钮以运行脚本。
-if __name__ == '__main__':
-    print_hi('PyCharm')
+print("数据预处理完成")
 
-# 访问 https://www.jetbrains.com/help/pycharm/ 获取 PyCharm 帮助
+model, label_map = train_model(d)
+print("训练完成")
+joblib.dump(model, "model.pkl")
+
+# model = joblib.load("model.pkl")
+X_train, y_train = prepare_data(d)
+print(predict(model, label_map, d))
+print(predict(model, label_map, val))
+# print(f"\n训练集准确率: {model.score(X_train, y_train):.5f}")
+# X_train, y_train = prepare_data(val)
+# print(f"\n训练集准确率: {model.score(X_train, y_train):.5f}")
+# model.predict()
+from src_predict import predictor
+suc = cnt = 0
+for file in Path("data/train").glob("*/*.jpg"):
+    cnt+=1
+    pcl,_=predictor(file)
+    acl = file.parents[0].name
+    if acl == pcl:
+        suc+=1
+
+print(f"预测准确率: {suc/cnt:.4f}")