Update

app.py

@@ -7,12 +7,16 @@ from collections import defaultdict
 from skimage.color import deltaE_ciede2000, rgb2lab
 import zipfile
 
-def DoG_filter(image, kernel_size=0, sigma=1.0, k_sigma=2.0, gamma=1.5):
-    g1 = cv2.GaussianBlur(image, (kernel_size, kernel_size), sigma)
-    g2 = cv2.GaussianBlur(image, (kernel_size, kernel_size), sigma * k_sigma)
+def DoG_filter(image, kernel_size, sigma, k_sigma, gamma):
+    if kernel_size == 0:
+        kernel_size = None  # OpenCVにカーネルサイズを自動で決定させる
+    else:
+        kernel_size = (kernel_size, kernel_size)
+    g1 = cv2.GaussianBlur(image, kernel_size, sigma)
+    g2 = cv2.GaussianBlur(image, kernel_size, sigma * k_sigma)
     return g1 - gamma * g2
 
-def XDoG_filter(image, kernel_size=0, sigma=1.4, k_sigma=1.6, epsilon=0, phi=10, gamma=0.98):
+def XDoG_filter(image, kernel_size, sigma, k_sigma, epsilon, phi, gamma):
     epsilon /= 255
     dog = DoG_filter(image, kernel_size, sigma, k_sigma, gamma)
     dog /= dog.max()
@@ -20,24 +24,10 @@ def XDoG_filter(image, kernel_size=0, sigma=1.4, k_sigma=1.6, epsilon=0, phi=10,
     e[e >= 1] = 1
     return (e * 255).astype('uint8')
 
-def binarize_image(image):
-    _, binarized = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
-    return binarized
-
-
-def process_XDoG(image_path):
-    kernel_size=0
-    sigma=1.4
-    k_sigma=1.6
-    epsilon=0
-    phi=10
-    gamma=0.98
-
-    image = cv2.imread(image_path, cv2.IMREAD_GRAYSCALE)
+def process_XDoG(image, kernel_size, sigma, k_sigma, epsilon, phi, gamma):
+    image = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2GRAY)
     xdog_image = XDoG_filter(image, kernel_size, sigma, k_sigma, epsilon, phi, gamma)
-    binarized_image = binarize_image(xdog_image)
-    final_image = Image.fromarray(binarized_image)
-    return final_image
+    return Image.fromarray(xdog_image).convert('L')
 
 
 
@@ -161,7 +151,14 @@ class webui:
     def __init__(self):
         self.demo = gr.Blocks()
 
-    def main(self, image_path):
+    def main(self, image_path, kernel_size, sigma, k_sigma, epsilon, phi, gamma):
+        kernel_size = int(kernel_size)
+        sigma = float(sigma)
+        k_sigma = float(k_sigma)
+        epsilon = int(epsilon)
+        phi = int(phi)
+        gamma = float(gamma)
+        
         image = Image.open(image_path).convert('RGBA')
         #拡張子を取り除いたファイル名を取得
         image_name = os.path.splitext(image_path)[0]
@@ -169,7 +166,7 @@ class webui:
         image.save(image_path)
         image = Image.open(image_path).convert('RGBA')
         rgb_image = image.convert('RGB')
-        lineart = process_XDoG(image_path).convert('L')
+        lineart = process_XDoG(image, kernel_size, sigma, k_sigma, epsilon, phi, gamma).convert('L')
         replace_color_image = process_image(rgb_image, lineart).convert('RGBA')
         
         if alpha:
@@ -197,19 +194,22 @@ class webui:
             with gr.Row():
                 with gr.Column():
                     input_image = gr.Image(type='filepath', image_mode="RGBA", label="Original Image")
+                    kernel_size = gr.Dropdown(choices=[0, 1, 3, 5, 7, 9], value=0, label="カーネルサイズ (kernel_size)", description="ガウシアンフィルタのカーネルサイズです。0を選択するとOpenCVが自動でカーネルサイズを決定します。奇数値を選ぶとより鋭い線が得られます。")
+                    sigma = gr.Slider(0.1, 10.0, step=0.1, value=1.4, label="シグマ (sigma)", description="ガウシアンフィルタの標準偏差。ぼかしの程度を制御します。")
+                    k_sigma = gr.Slider(1.0, 3.0, step=0.1, value=1.6, label="k_シグマ (k_sigma)", description="二つ目のガウシアンフィルタのシグマの倍率。線の太さが影響を受けます。")
+                    epsilon = gr.Slider(-10, 10, value=0, label="イプシロン (epsilon)", description="XDoGフィルターの閾値パラメータ。線画の明瞭度やコントラストを調整します。")
+                    phi = gr.Slider(1, 100, value=10, label="ファイ (phi)", description="タンジェントハイパーボリック関数の係数。閾値を超えた部分の鮮明さを高めます。")
+                    gamma = gr.Slider(0.5, 1.5, step=0.1, value=0.98, label="ガンマ (gamma)", description="DoGフィルタの強度調整パラメータ。主に線の明るさや暗さを調節します。")
                     submit = gr.Button(value="Start")
-                with gr.Row():
-                    with gr.Column():
-                        with gr.Tab("output"):
-                            output_0 = gr.Gallery(format="png")
-                        output_file = gr.File()
+                with gr.Column():
+                    output_image = gr.Image(label="Processed Image")
+
             submit.click(
                 self.main, 
-                inputs=[input_image], 
-                outputs=[output_0, output_file]
+                inputs=[input_image, kernel_size, sigma, k_sigma, epsilon, phi, gamma],
+                outputs=[output_image]
             )
 
-        self.demo.queue()
         self.demo.launch(share=share)
 
 if __name__ == "__main__":