req

dataset_aug.py

@@ -16,9 +16,12 @@ from tqdm import tqdm
 import argparse
 
 def get_average_color(image):
-    """画像の平均色を計算する"""
+    """画像の平均色を取得"""
+    # RGBAの場合はRGBに変換
+    if image.mode == 'RGBA':
+        image = image.convert('RGB')
+    
     stat = ImageStat.Stat(image)
-    # 平均色を取得（RGB）
     r, g, b = map(int, stat.mean)
     return (r, g, b)
 

requirements.txt

@@ -15,3 +15,4 @@ accelerate==0.13.1
 scipy==1.9.3
 pybooru==4.2.2
 webdataset==0.2.86
+opencv-python==4.11.0.86

test_app.py

@@ -19,13 +19,19 @@ def process_multiple_images(
     source_is_edge_mode_fill,
     target_is_avg_color_fill,
     target_is_edge_mode_fill,
-    expand_to_long_side
+    expand_to_long_side,
+    progress=gr.Progress()
 ):
     result_source_images = []
     result_target_images = []
     
+    progress(0, desc="処理を開始します...")
+    
     # 各画像ペアに対して処理を実行
-    for source_path, target_path in zip(source_images, target_images):
+    total_pairs = len(source_images)
+    for idx, (source_path, target_path) in enumerate(zip(source_images, target_images), 1):
+        progress(idx/total_pairs, desc=f"画像ペア {idx}/{total_pairs} を処理中...")
+        
         # PILイメージとして読み込み
         source_img = Image.open(source_path.name)
         target_img = Image.open(target_path.name)
@@ -51,7 +57,8 @@ def process_multiple_images(
         result_source_images.extend(aug_sources)
         result_target_images.extend(aug_targets)
     
-    return result_source_images, result_target_images
+    progress(1, desc="処理が完了しました")
+    return result_source_images, result_target_images, "処理が完了しました！"
 
 def update_source_preview(source_files):
     preview_images = []
@@ -181,7 +188,7 @@ def test_process_image_pair_with_expand_to_long_side():
 
 # Gradioインターフェースの作成
 with gr.Blocks() as demo:
-    gr.Markdown("# データ拡張テスト")
+    gr.Markdown("# Pair Image Augmentation Test")
     gr.Markdown("Code : https://github.com/Yeq6X/pair-images-aug")
     
     with gr.Row():
@@ -219,8 +226,6 @@ with gr.Blocks() as demo:
                 show_label=True,
                 object_fit="contain",
                 columns=4,
-                height=300,
-                preview=True,
             )
             with gr.Row():
                 gr.Markdown("### scribble_xdogで変換")
@@ -232,16 +237,12 @@ with gr.Blocks() as demo:
                 show_label=True,
                 object_fit="contain",
                 columns=4,
-                height=300,
-                preview=True
             )           
             
         # 右側のカラム（Target画像と出力）
         with gr.Column():
             # パラメータ設定部分
-            with gr.Group():
-                gr.Markdown("### パラメータ設定")
-                
+            with gr.Accordion("パラメータ設定", open=False):
                 # パラメータ操作ボタン
                 with gr.Row():
                     randomize_btn = gr.Button("🎲 ランダム設定", variant="secondary")
@@ -316,14 +317,22 @@ with gr.Blocks() as demo:
             
             process_btn = gr.Button("処理開始", variant="primary")
 
+            # ログ表示用のテキストボックスを追加
+            log_output = gr.Textbox(
+                label="処理ログ",
+                value="",
+                lines=1,
+                max_lines=1,
+                interactive=False,
+                visible=False
+            )
+
             # 結果表示
             result_source_gallery = gr.Gallery(
                 label="生成結果 (Source)",
                 show_label=True,
                 object_fit="contain",
                 columns=4,
-                height=250,
-                preview=True,
                 type="pil"
             )
             result_target_gallery = gr.Gallery(
@@ -331,11 +340,9 @@ with gr.Blocks() as demo:
                 show_label=True,
                 object_fit="contain",
                 columns=4,
-                height=250,
-                preview=True,
                 type="pil"
             )
-    
+        
     # イベントハンドラ
     source_files.change(
         fn=update_source_preview,
@@ -404,7 +411,7 @@ with gr.Blocks() as demo:
             target_is_edge_mode_fill,
             expand_to_long_side
         ],
-        outputs=[result_source_gallery, result_target_gallery]
+        outputs=[result_source_gallery, result_target_gallery, log_output]
     )
 
 if __name__ == "__main__":