Update app.py

app.py

@@ -1,278 +1,224 @@
 import gradio as gr
-import polars as pl
+import pandas as pd
 import os
 import shutil
+import matplotlib.pyplot as plt
 import numpy as np
-from PyQt6.QtWidgets import QApplication
-from PyQt6 import QtCore
-import pyqtgraph as pg
-from pyqtgraph.exporters import ImageExporter
+import japanize_matplotlib
 
-class DataProcessor:
-    def __init__(self, bodypart_names, x_max, y_max):
-        # 余分な空白を除去してリスト化
-        self.bodypart_names = [name.strip() for name in bodypart_names.split(',')]
-        self.x_max = x_max
-        self.y_max = y_max
-        self.output_folder = 'output_plots'
-        if not os.path.exists(self.output_folder):
-            os.makedirs(self.output_folder)
-
-    def process_csv(self, file_path):
-        # CSVをpolarsで読み込み（ヘッダーはなし）
-        df_raw = pl.read_csv(file_path, has_header=False)
-        # 2行分のヘッダー（インデックス1,2）を取得し、最初の列は除外する
-        header1 = df_raw.row(1)[1:]
-        header2 = df_raw.row(2)[1:]
-        new_columns = [f"{h1}|{h2}" for h1, h2 in zip(header1, header2)]
-        # データ部分はインデックス3以降（0-indexed）とし、先頭列を削除
-        df_data = df_raw.slice(3)
-        first_col = df_data.columns[0]
-        df_data = df_data.drop(first_col)
-        df_data.columns = new_columns
-
-        # likelihood列のみ抽出
-        df_likelihood = self.extract_likelihood(df_data)
-        # likelihood列を除去したデータ
-        df_no_likelihood = self.remove_likelihood(df_data)
-        # 付属肢名の置換（左側の名前を mapping で変更）
-        df_renamed = self.rename_bodyparts(df_no_likelihood)
-        return df_renamed, df_likelihood
-
-    def remove_likelihood(self, df):
-        # 列名が "bodypart|likelihood" となっている列を除外
-        new_cols = [col for col in df.columns if col.split("|")[1] != "likelihood"]
-        return df.select(new_cols)
-
-    def rename_bodyparts(self, df):
-        cols = df.columns
-        current_names = []
-        for col in cols:
-            bp = col.split("|")[0]
-            if bp not in current_names:
-                current_names.append(bp)
-        if len(self.bodypart_names) != len(current_names):
-            raise ValueError("The length of bodypart_names must be equal to the number of bodyparts.")
-        mapping = dict(zip(current_names, self.bodypart_names))
-        new_cols = {col: f"{mapping[col.split('|')[0]]}|{col.split('|')[1]}" for col in cols}
-        return df.rename(new_cols)
-
-    def extract_likelihood(self, df):
-        # likelihood列のみを抽出
-        likelihood_cols = [col for col in df.columns if col.split("|")[1] == "likelihood"]
-        df_likelihood = df.select(likelihood_cols)
-        current_names = []
-        for col in likelihood_cols:
-            bp = col.split("|")[0]
-            if bp not in current_names:
-                current_names.append(bp)
-        if len(self.bodypart_names) != len(current_names):
-            raise ValueError("The length of bodypart_names must be equal to the number of bodyparts.")
-        mapping = dict(zip(current_names, self.bodypart_names))
-        new_cols = {col: f"{mapping[col.split('|')[0]]}|{col.split('|')[1]}" for col in likelihood_cols}
-        return df_likelihood.rename(new_cols)
-
-    def get_bodyparts(self, df):
-        bodyparts = []
-        for col in df.columns:
-            bp = col.split("|")[0]
-            if bp not in bodyparts:
-                bodyparts.append(bp)
-        return bodyparts
-
-    def plot_scatter(self, df):
-        image_paths = []
-        bodyparts = self.get_bodyparts(df)
-        app = QApplication.instance()
-        if app is None:
-            app = QApplication([])
-
-        # 個別の散布図を作成
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                x = np.array(df[f"{bodypart}|x"].to_list(), dtype=float)
-                y = np.array(df[f"{bodypart}|y"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            pw = pg.PlotWidget(title=f'トラッキングの座標({bodypart})')
-            pw.setLabel('bottom', 'X Coordinate(pixel)')
-            pw.setLabel('left', 'Y Coordinate(pixel)')
-            pw.setXRange(0, self.x_max)
-            pw.setYRange(0, self.y_max)
-            pw.invertY(True)
-            color = pg.intColor(i, len(bodyparts))
-            # 散布図アイテムの追加
-            scatter = pg.ScatterPlotItem(x=x, y=y, pen=pg.mkPen(color=color), symbol='o', brush=color)
-            pw.addItem(scatter)
-            # 始点を黒丸でハイライトし、"Start"テキストを追加
-            if len(x) > 0:
-                scatter_start = pg.ScatterPlotItem(x=[x[0]], y=[y[0]], pen=pg.mkPen(color='k'), symbol='o', size=10, brush='k')
-                pw.addItem(scatter_start)
-                text = pg.TextItem("Start", color='k')
-                text.setPos(x[0], y[0])
-                pw.addItem(text)
-            # PNGにエクスポート
-            exporter = ImageExporter(pw.plotItem)
-            filename = os.path.join(self.output_folder, f"{bodypart}.png")
-            exporter.export(filename)
-            image_paths.append(filename)
-
-        # 全付属肢の散布図を作成
-        pw_all = pg.PlotWidget(title='トラッキングの座標(全付属肢)')
-        pw_all.setLabel('bottom', 'X Coordinate(pixel)')
-        pw_all.setLabel('left', 'Y Coordinate(pixel)')
-        pw_all.setXRange(0, self.x_max)
-        pw_all.setYRange(0, self.y_max)
-        pw_all.invertY(True)
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                x = np.array(df[f"{bodypart}|x"].to_list(), dtype=float)
-                y = np.array(df[f"{bodypart}|y"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            color = pg.intColor(i, len(bodyparts))
-            scatter = pg.ScatterPlotItem(x=x, y=y, pen=pg.mkPen(color=color), symbol='o', brush=color)
-            pw_all.addItem(scatter)
-        exporter_all = ImageExporter(pw_all.plotItem)
-        filename_all = os.path.join(self.output_folder, "all_plot.png")
-        exporter_all.export(filename_all)
-        image_paths.append(filename_all)
-        return image_paths
 
-    def plot_trajectories(self, df):
-        image_paths = []
-        bodyparts = self.get_bodyparts(df)
-        app = QApplication.instance()
-        if app is None:
-            app = QApplication([])
-
-        # 個別の軌跡図を作成
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                x = np.array(df[f"{bodypart}|x"].to_list(), dtype=float)
-                y = np.array(df[f"{bodypart}|y"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            pw = pg.PlotWidget(title=f'トラッキングの座標({bodypart})')
-            pw.setLabel('bottom', 'Frames')
-            pw.setLabel('left', 'Coordinate(pixel)')
-            pen_x = pg.mkPen(color=pg.intColor(i, len(bodyparts)), style=QtCore.Qt.PenStyle.DashLine)
-            pen_y = pg.mkPen(color=pg.intColor(i, len(bodyparts)))
-            pw.plot(x, pen=pen_x, name=f"{bodypart}(x座標)")
-            pw.plot(y, pen=pen_y, name=f"{bodypart}(y座標)")
-            exporter = ImageExporter(pw.plotItem)
-            filename = os.path.join(self.output_folder, f"{bodypart}_trajectories.png")
-            exporter.export(filename)
-            image_paths.append(filename)
-
-        # 全付属肢の軌跡図を作成
-        pw_all = pg.PlotWidget(title='トラッキングの座標(全付属肢)')
-        pw_all.setLabel('bottom', 'Frames')
-        pw_all.setLabel('left', 'Coordinate(pixel)')
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                x = np.array(df[f"{bodypart}|x"].to_list(), dtype=float)
-                y = np.array(df[f"{bodypart}|y"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            pen_x = pg.mkPen(color=pg.intColor(i, len(bodyparts)), style=QtCore.Qt.PenStyle.DashLine)
-            pen_y = pg.mkPen(color=pg.intColor(i, len(bodyparts)))
-            pw_all.plot(x, pen=pen_x, name=f"{bodypart}(x座標)")
-            pw_all.plot(y, pen=pen_y, name=f"{bodypart}(y座標)")
-        exporter_all = ImageExporter(pw_all.plotItem)
-        filename_all = os.path.join(self.output_folder, "all_trajectories.png")
-        exporter_all.export(filename_all)
-        image_paths.append(filename_all)
-        return image_paths
-
-    def plot_likelihood(self, df_likelihood):
-        image_paths = []
-        bodyparts = self.get_bodyparts(df_likelihood)
-        app = QApplication.instance()
-        if app is None:
-            app = QApplication([])
-
-        # 付属肢ごとの尤度グラフを作成
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                likelihood = np.array(df_likelihood[f"{bodypart}|likelihood"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            pw = pg.PlotWidget(title=f'フレーム別の尤度 ({bodypart})')
-            pw.setLabel('bottom', 'Frames')
-            pw.setLabel('left', '尤度')
-            pw.setYRange(0, 1.0)
-            color = pg.intColor(i, len(bodyparts))
-            pw.plot(likelihood, pen=pg.mkPen(color=color), name=bodypart)
-            exporter = ImageExporter(pw.plotItem)
-            filename = os.path.join(self.output_folder, f"{bodypart}_likelihood.png")
-            exporter.export(filename)
-            image_paths.append(filename)
-
-        # 全付属肢の尤度グラフを作成
-        pw_all = pg.PlotWidget(title='フレーム別の尤度 (全付属肢)')
-        pw_all.setLabel('bottom', 'Frames')
-        pw_all.setLabel('left', '尤度')
-        pw_all.setYRange(0, 1.0)
-        for i, bodypart in enumerate(bodyparts):
-            try:
-                likelihood = np.array(df_likelihood[f"{bodypart}|likelihood"].to_list(), dtype=float)
-            except Exception as e:
-                continue
-            color = pg.intColor(i, len(bodyparts))
-            pw_all.plot(likelihood, pen=pg.mkPen(color=color), name=bodypart)
-        exporter_all = ImageExporter(pw_all.plotItem)
-        filename_all = os.path.join(self.output_folder, "likelihood_plot.png")
-        exporter_all.export(filename_all)
-        image_paths.append(filename_all)
-        return image_paths
+class DataProcessor:
+   def __init__(self, bodypart_names, x_max, y_max):
+      self.bodypart_names = bodypart_names.split(',')
+      self.x_max = x_max
+      self.y_max = y_max
+      self.output_folder = 'output_plots'
+
+   def process_csv(self, file_path):
+      df = pd.read_csv(file_path, header=[1, 2])
+      df_likelihood = self.extract_likelihood(df)
+      df = self.remove_first_column_and_likelihood(df)
+      df = self.rename_bodyparts(df)
+      return df, df_likelihood
+
+   def remove_first_column_and_likelihood(self, df):
+      df = df.drop(df.columns[0], axis=1)
+      df = df[df.columns.drop(list(df.filter(regex='likelihood')))]
+      return df
+
+   def rename_bodyparts(self, df):
+      current_names = df.columns.get_level_values(0).unique()
+      if len(self.bodypart_names) != len(current_names):
+         raise ValueError(
+             "The length of bodypart_names must be equal to the number of bodyparts.")
+      mapping = dict(zip(current_names, self.bodypart_names))
+      new_columns = [(mapping[col[0]], col[1]) if col[0]
+                     in mapping else col for col in df.columns]
+      df.columns = pd.MultiIndex.from_tuples(new_columns)
+      return df
+
+   def extract_likelihood(self, df):
+      # likelihood列のみを抽出する
+      df = df[df.columns[df.columns.get_level_values(1) == 'likelihood']]
+      df.drop(df.columns[0], axis=1)
+      current_names = df.columns.get_level_values(0).unique()
+      mapping = dict(zip(current_names, self.bodypart_names))
+      new_columns = [(mapping[col[0]], col[1]) if col[0]
+                     in mapping else col for col in df.columns]
+      df.columns = pd.MultiIndex.from_tuples(new_columns)
+
+      return df
+
+   def plot_scatter(self, df):
+      if not os.path.exists(self.output_folder):
+         os.makedirs(self.output_folder)
+      return self.plot_scatter_fixed(df, self.output_folder, self.x_max, self.y_max)
+
+   def plot_scatter_fixed(self, df, output_folder, x_max, y_max):
+      image_paths = []
+      bodyparts = df.columns.get_level_values(0).unique()
+      colors = plt.cm.rainbow(np.linspace(0, 1, len(bodyparts)))
+      for i, bodypart in enumerate(bodyparts):
+         x = df[bodypart]['x'].values
+         y = df[bodypart]['y'].values
+         plt.figure(figsize=(8, 6))
+         plt.scatter(x, y, color=colors[i], label=bodypart)
+         plt.scatter(x[0], y[0], color='black', marker='o', s=100)
+         plt.text(x[0], y[0], ' Start', color='black',
+                  fontsize=12, verticalalignment='bottom')
+         plt.xlim(0, x_max)
+         plt.ylim(0, y_max)
+         plt.gca().invert_yaxis()
+         plt.title(f'トラッキングの座標({bodypart})')
+         plt.xlabel('X Coordinate(pixel)')
+         plt.ylabel('Y Coordinate(pixel)')
+         plt.legend(loc='upper right')
+         plt.savefig(f'{output_folder}/{bodypart}.png')
+         image_paths.append(f'{output_folder}/{bodypart}.png')
+         plt.close()
+
+      plt.figure(figsize=(8, 6))
+      for i, bodypart in enumerate(bodyparts):
+         x = df[bodypart]['x'].values
+         y = df[bodypart]['y'].values
+         plt.scatter(x, y, color=colors[i], label=bodypart)
+      plt.xlim(0, x_max)
+      plt.ylim(0, y_max)
+      plt.gca().invert_yaxis()
+      plt.title('トラッキングの座標(全付属肢)')
+      plt.xlabel('X Coordinate(pixel)')
+      plt.ylabel('Y Coordinate(pixel)')
+      plt.legend(loc='upper right')
+      plt.savefig(f'{output_folder}/all_plot.png')
+      image_paths.append(f'{output_folder}/all_plot.png')
+      plt.close()
+
+      return image_paths
+
+   def plot_trajectories(self, df):
+      image_paths = []
+      bodyparts = df.columns.get_level_values(0).unique()
+      colors = plt.cm.rainbow(np.linspace(0, 1, len(bodyparts)))
+
+      for i, bodypart in enumerate(bodyparts):
+         x = df[bodypart]['x'].values
+         y = df[bodypart]['y'].values
+         plt.figure(figsize=(8, 6))
+         plt.plot(x, color=colors[i], label=bodypart +
+                  "(x座標)", linestyle='dashed')
+         plt.plot(y, color=colors[i], label=bodypart + "(y座標)")
+         plt.title(f'トラッキングの座標({bodypart})')
+         plt.xlabel('Frames')
+         plt.ylabel('Coordinate(pixel)')
+         plt.legend(loc='upper right')
+         plt.savefig(f'{self.output_folder}/{bodypart}_trajectories.png')
+         image_paths.append(
+             f'{self.output_folder}/{bodypart}_trajectories.png')
+         plt.close()
+
+      plt.figure(figsize=(8, 6))
+      for i, bodypart in enumerate(bodyparts):
+         x = df[bodypart]['x'].values
+         y = df[bodypart]['y'].values
+         plt.plot(x, color=colors[i], label=bodypart +
+                  "(x座標)", linestyle='dashed')
+         plt.plot(y, color=colors[i], label=bodypart + "(y座標)")
+      plt.title(f'トラッキングの座標({bodypart})')
+      plt.xlabel('Frames')
+      plt.ylabel('Coordinate(pixel)')
+      plt.legend(loc='upper right')
+      plt.savefig(f'{self.output_folder}/all_trajectories.png')
+      image_paths.append(f'{self.output_folder}/all_trajectories.png')
+      plt.close()
+      return image_paths
+
+   def plot_likelihood(self, likelihood_df):
+      image_paths = []
+      plt.ylim(0, 1.0)
+      bodyparts = likelihood_df.columns.get_level_values(0).unique()
+      colors = plt.cm.rainbow(np.linspace(0, 1, len(bodyparts)))
+
+      # 付属肢ごとに尤度をプロット
+      for i, bodypart in enumerate(bodyparts):
+         plt.figure(figsize=(8, 6))
+         plt.ylim(0, 1.0)
+         plt.plot(likelihood_df[bodypart], color=colors[i], label=bodypart)
+         plt.xlabel('Frames')
+         plt.ylabel('尤度')
+         plt.title('フレーム別の尤度')
+         # 凡例を右上の外側に表示
+         plt.legend(bbox_to_anchor=(1.05, 1),
+                    loc='upper left', borderaxespad=0)
+         # 凡例がはみ出さないようにレイアウトを調整
+         plt.tight_layout()
+         plt.savefig(f'{self.output_folder}/{bodypart}_likelihood.png')
+         image_paths.append(f'{self.output_folder}/{bodypart}_likelihood.png')
+         plt.close()
+
+      # 全ての付属肢の尤度をプロット
+      plt.figure(figsize=(8, 6))
+      plt.ylim(0, 1.0)
+      for i, column in enumerate(likelihood_df.columns):
+         plt.plot(likelihood_df[column], color=colors[i], label=column[0])
+      plt.xlabel('Frames')
+      plt.ylabel('尤度')
+      plt.title('フレーム別の尤度')
+      # 凡例を右上の外側に表示
+      plt.legend(bbox_to_anchor=(1.05, 1), loc='upper left', borderaxespad=0)
+      # 凡例がはみ出さないようにレイアウトを調整
+      plt.tight_layout()
+      plt.savefig(f'{self.output_folder}/likelihood_plot.png')
+      plt.close()
+      image_paths.append(f'{self.output_folder}/likelihood_plot.png')
+      return image_paths
+
+# 以下のGradioInterfaceクラスとメイン実行部分は変更なし
 
 class GradioInterface:
-    def __init__(self):
-        self.interface = gr.Interface(
-            fn=self.process_and_plot,
-            inputs=[
-                gr.File(label="CSVファイルをドラッグ＆ドロップ"),
-                gr.Textbox(
-                    label="付属肢の名前（カンマ区切り）",
-                    value="指節1, 指節2, 指節3, 指節4, 指節5, 指節6, 指節7, 指節8, 指節9, 指節10, 指節11, 指節12, 指節13, 指節14, 触角(左), 触角(右), 頭部, 腹尾節"
-                ),
-                gr.Number(label="X軸の最大値", value=1920),
-                gr.Number(label="Y軸の最大値", value=1080),
-                gr.CheckboxGroup(
-                    label="プロットするグラフを選択",
-                    choices=["散布図", "軌跡図", "尤度グラフ"],
-                    value=["散布図", "軌跡図", "尤度グラフ"],
-                    type="value"
-                )
-            ],
-            outputs=[
-                gr.Gallery(label="散布図"),
-                gr.File(label="ZIPダウンロード")
-            ],
-            title="DeepLabCutグラフ出力ツール",
-            description="CSVファイルからグラフを作成します。"
-        )
-
-    def process_and_plot(self, file, bodypart_names, x_max, y_max, graph_choices):
-        processor = DataProcessor(bodypart_names, x_max, y_max)
-        df, df_likelihood = processor.process_csv(file.name)
-
-        all_image_paths = []
-        if "散布図" in graph_choices:
-            all_image_paths += processor.plot_scatter(df)
-        if "軌跡図" in graph_choices:
-            all_image_paths += processor.plot_trajectories(df)
-        if "尤度グラフ" in graph_choices:
-            all_image_paths += processor.plot_likelihood(df_likelihood)
-
-        shutil.make_archive(processor.output_folder, 'zip', processor.output_folder)
-        return all_image_paths, processor.output_folder + '.zip'
-
-    def launch(self):
-        self.interface.launch()
+   def __init__(self):
+      self.interface = gr.Interface(
+          fn=self.process_and_plot,
+          inputs=[
+              gr.File(label="CSVファイルをドラッグ＆ドロップ"),
+              gr.Textbox(label="付属肢の名前（カンマ区切り）",
+                         value="指節1, 指節2, 指節3, 指節4, 指節5, 指節6, 指節7, 指節8, 指節9,指節10, 指節11, 指節12, 指節13, 指節14, 触角(左), 触角(右), 頭部, 腹尾節"),
+              gr.Number(label="X軸の最大値", value=1920),
+              gr.Number(label="Y軸の最大値", value=1080),
+              gr.CheckboxGroup(
+                  label="プロットするグラフを選択",
+                  choices=["散布図", "軌跡図", "尤度グラフ"],
+                  value=["散布図", "軌跡図", "尤度グラフ"],
+                  type="value"
+              )
+          ],
+          outputs=[
+              gr.Gallery(label="散布図"),
+              gr.File(label="ZIPダウンロード")
+          ],
+          title="DeepLabCutグラフ出力ツール",
+          description="CSVファイルからグラフを作成します。"
+      )
+
+   def process_and_plot(self, file, bodypart_names, x_max, y_max, graph_choices):
+      processor = DataProcessor(bodypart_names, x_max, y_max)
+      df, df_likelihood = processor.process_csv(file.name)
+
+      all_image_paths = []
+      if "散布図" in graph_choices:
+         all_image_paths += processor.plot_scatter(df)
+      if "軌跡図" in graph_choices:
+         all_image_paths += processor.plot_trajectories(df)
+      if "尤度グラフ" in graph_choices:
+         all_image_paths += processor.plot_likelihood(df_likelihood)
+
+      shutil.make_archive(processor.output_folder,
+                          'zip', processor.output_folder)
+      return all_image_paths, processor.output_folder + '.zip'
+
+   def launch(self):
+      self.interface.launch()
 
 
 if __name__ == "__main__":
-    gradio_app = GradioInterface()
-    gradio_app.launch()
+   gradio_app = GradioInterface()
+   gradio_app.launch()