Update

app.py

@@ -31,18 +31,8 @@ def preprocess_image(image):
     image = image.astype(np.float32)
     return image
 
-def process_image(image, ort_sess, input_name, rating_tags, character_tags, general_tags):
+def process_image(prob, rating_tags, character_tags, general_tags):
     thresh = 0.35
-    try:
-        image = Image.open(image_path)
-        image = image.convert("RGB") if image.mode != "RGB" else image
-        image = preprocess_image(image)
-    except Exception as e:
-        print(f"画像を読み込めません: {image_path}, エラー: {e}")
-        return
-
-    img = np.array([image])
-    prob = ort_sess.run(None, {input_name: img})[0][0]  # ONNXモデルからの出力
 
     # NSFW/SFW判定
     tag_confidences = {tag: prob[i] for i, tag in enumerate(rating_tags)}
@@ -138,7 +128,11 @@ class webui:
         character_tags = [row[1] for row in rows if row[2] == "4"]
         general_tags = [row[1] for row in rows if row[2] == "0"]
 
-        NSFW_flag, IP_flag, tag_text = process_image(image, ort_sess, ort_sess.get_inputs()[0].name, rating_tags, character_tags, general_tags)
+
+        img = np.array([image])
+        prob = ort_sess.run(None, {ort_sess.get_inputs()[0].name: img})[0][0]  # ONNXモデルからの出力
+
+        NSFW_flag, IP_flag, tag_text = process_image(prob, rating_tags, character_tags, general_tags)
         return NSFW_flag, IP_flag, tag_text
 
 
@@ -147,7 +141,7 @@ class webui:
             with gr.Row():
                 with gr.Column():
                     input_image = gr.Image(type='filepath', label="Analysis Image")
-                    model_id = gr.Textbox(label="NSFW Flag", value="SmilingWolf/wd-vit-tagger-v3")
+                    model_id = gr.Textbox(label="MODEL ID", value="SmilingWolf/wd-vit-tagger-v3")
                     output_0 = gr.Textbox(label="NSFW Flag")
                     output_1 = gr.Textbox(label="IP Flag")
                     output_2 = gr.Textbox(label="Tags")