Update Requirements

app.py

@@ -3,6 +3,8 @@ import pandas as pd
 from bokeh.plotting import figure
 from bokeh.models import ColumnDataSource
 from bokeh.palettes import Category10
+from sklearn.decomposition import PCA
+from sklearn.manifold import TSNE
 
 TOOLTIPS = """
 <div>
@@ -86,64 +88,78 @@ if __name__ == "__main__":
     config_style()
 
     # --- Primer gráfico: datos de Donut ---
-    df_donut_pca = pd.read_csv("data/data_donut_pca.csv")
-    df_donut_tsne = pd.read_csv("data/data_donut_tsne.csv")
-
-    # Desplegable para seleccionar visualización
+    # Se asume que "embeddings_donut.csv" contiene las columnas "dim_0", "dim_1", ..., "dim_N", además de "label" e "img"
+    df_donut = pd.read_csv("data/donut_de_Rodrigo_merit_secret_all_embeddings.csv")
+    
+    # Selección de visualización
     donut_mode = st.selectbox(
         "Seleccione visualización para Donut:",
         options=["PCA", "t-SNE"]
     )
     
-    # Escoger el DataFrame según la selección
+    # Extraer columnas de embedding (aquellas que empiezan con "dim_")
+    embedding_cols = [col for col in df_donut.columns if col.startswith("dim_")]
+    all_embeddings = df_donut[embedding_cols].values
+    
     if donut_mode == "PCA":
-        current_df_donut = df_donut_pca
+        pca = PCA(n_components=2)
+        reduced = pca.fit_transform(all_embeddings)
     else:
-        current_df_donut = df_donut_tsne
-
-    unique_labels = current_df_donut['label'].unique().tolist()
+        tsne = TSNE(n_components=2, random_state=42, perplexity=30, learning_rate=200)
+        reduced = tsne.fit_transform(all_embeddings)
+    
+    # Añadir las coordenadas resultantes al DataFrame
+    df_donut['x'] = reduced[:, 0]
+    df_donut['y'] = reduced[:, 1]
+    
+    unique_labels = df_donut['label'].unique().tolist()
     plot_placeholder = st.empty()
-
+    
     # Mostrar gráfico inicial con todas las etiquetas
-    render_plot(unique_labels, current_df_donut, plot_placeholder)
-
+    render_plot(unique_labels, df_donut, plot_placeholder)
+    
     # Desplegable para filtrar etiquetas
     selected_labels = st.multiselect(
         "Seleccione subsets para visualizar (Donut):",
         options=unique_labels,
         default=unique_labels
     )
-
-    render_plot(selected_labels, current_df_donut, plot_placeholder)
-
+    render_plot(selected_labels, df_donut, plot_placeholder)
+    
     # --- Segundo gráfico: datos de Idefics2 ---
     st.markdown('<h2 class="sub-title">Idefics2</h2>', unsafe_allow_html=True)
-
-    df_idefics2_pca = pd.read_csv("data/data_idefics2_pca.csv")
-    df_idefics2_tsne = pd.read_csv("data/data_idefics2_tsne.csv")
-
-    # Desplegable para seleccionar visualización para Idefics2
+    
+    # Se asume que "embeddings_idefics2.csv" tiene la misma estructura
+    df_idefics2 = pd.read_csv("data/embeddings_idefics2.csv")
+    
     idefics2_mode = st.selectbox(
         "Seleccione visualización para Idefics2:",
         options=["PCA", "t-SNE"],
         key="idefics2_mode"
     )
     
+    embedding_cols2 = [col for col in df_idefics2.columns if col.startswith("dim_")]
+    all_embeddings2 = df_idefics2[embedding_cols2].values
+    
     if idefics2_mode == "PCA":
-        current_df_idefics2 = df_idefics2_pca
+        pca2 = PCA(n_components=2)
+        reduced2 = pca2.fit_transform(all_embeddings2)
     else:
-        current_df_idefics2 = df_idefics2_tsne
-
-    unique_labels2 = current_df_idefics2['label'].unique().tolist()
+        tsne2 = TSNE(n_components=2, random_state=42, perplexity=30, learning_rate=200)
+        reduced2 = tsne2.fit_transform(all_embeddings2)
+    
+    df_idefics2['x'] = reduced2[:, 0]
+    df_idefics2['y'] = reduced2[:, 1]
+    
+    unique_labels2 = df_idefics2['label'].unique().tolist()
     plot_placeholder2 = st.empty()
-
-    render_plot(unique_labels2, current_df_idefics2, plot_placeholder2)
-
+    
+    render_plot(unique_labels2, df_idefics2, plot_placeholder2)
+    
     selected_labels2 = st.multiselect(
         "Seleccione subsets para visualizar (Idefics2):",
         options=unique_labels2,
         default=unique_labels2,
         key="idefics2"
     )
-
-    render_plot(selected_labels2, current_df_idefics2, plot_placeholder2)
+    render_plot(selected_labels2, df_idefics2, plot_placeholder2)