Upload app.py with huggingface_hub

app.py

@@ -129,30 +129,41 @@ def get_training_embeddings_sample(n=100000):
     """Load a sample of training data for recommendations."""
     import pandas as pd
     from sklearn.model_selection import train_test_split
-    
+
     data = pd.read_csv('xkcd_scaled_data_Final.txt', nrows=n)
     data = data.dropna(subset=['name']).reset_index(drop=True)
     data['name_clean'] = data['name'].apply(clean_color_name)
     data = data[data['name_clean'].str.len() > 0].reset_index(drop=True)
-    
+
     indices = np.arange(len(data))
     train_idx, _ = train_test_split(indices, test_size=0.25, random_state=42)
-    
+
     train_names = data.loc[train_idx, 'name_clean'].tolist()
     train_rgbs = data.loc[train_idx, ['red', 'green', 'blue']].values
-    
+
     # Build embeddings
     X = np.zeros((len(train_names), MAX_TOKENS, VEC_SIZE), dtype=np.float32)
     for i, name in enumerate(train_names):
         tokens = name.split()
         for j, token in enumerate(tokens[:MAX_TOKENS]):
             X[i, j] = fasttext.wv[token]
-    
+
     return X, train_names, train_rgbs
 
+# Try to load training data for recommendations (optional)
 print("Loading training data for recommendations...")
-X_train, train_names, train_rgbs = get_training_embeddings_sample(100000)
-print(f"Loaded {len(train_names)} training samples for recommendations")
+try:
+    X_train, train_names, train_rgbs = get_training_embeddings_sample(100000)
+    print(f"Loaded {len(train_names)} training samples for recommendations")
+    RECOMMENDATIONS_ENABLED = True
+except FileNotFoundError:
+    print("Warning: xkcd_scaled_data_Final.txt not found. Recommendations disabled.")
+    X_train, train_names, train_rgbs = None, None, None
+    RECOMMENDATIONS_ENABLED = False
+except Exception as e:
+    print(f"Warning: Failed to load training data: {e}. Recommendations disabled.")
+    X_train, train_names, train_rgbs = None, None, None
+    RECOMMENDATIONS_ENABLED = False
 
 # ── Color Space Conversions ──────────────────────────────────────────────────
 def rgb_to_hex(r, g, b):
@@ -215,11 +226,13 @@ def predict_color(name):
     return r, g, b
 
 def cosine_similarity(query_vec, reference_vecs=X_train):
+    if not RECOMMENDATIONS_ENABLED or reference_vecs is None:
+        return np.array([])
     m, n = query_vec.shape
     test_new = query_vec.reshape(m * n)
     word_mag = np.linalg.norm(test_new)
     if word_mag == 0:
-        return np.arange(len(reference_vecs))
+        return np.array([])
     
     p, q, r = reference_vecs.shape
     ref_new = reference_vecs.reshape(p, q * r)
@@ -230,6 +243,8 @@ def cosine_similarity(query_vec, reference_vecs=X_train):
 
 def get_recommendations(query_embedding, top_k=5):
     """Get top-k similar colors from training data."""
+    if not RECOMMENDATIONS_ENABLED:
+        return []
     indices = cosine_similarity(query_embedding)
     results = []
     for idx in indices[:top_k]:
@@ -266,26 +281,29 @@ def process_color_name(color_name):
     query_emb = get_embedding(color_name)
     recommendations = get_recommendations(query_emb, top_k=5)
     
-    rec_html = '<div style="display: flex; flex-direction: column; gap: 8px;">'
-    for name, rec_hex, rec_rgb in recommendations:
-        rec_html += f"""
-        <div style="display: flex; align-items: center; padding: 10px; background: #f9f9f9; border-radius: 8px; border: 1px solid #eee;">
-            <div style="
-                width: 50px; 
-                height: 50px; 
-                background-color: {rec_hex}; 
-                border: 1px solid #ddd; 
-                border-radius: 6px;
-                margin-right: 16px;
-                flex-shrink: 0;
-            "></div>
-            <div>
-                <div style="font-weight: 600; font-size: 14px;">{name}</div>
-                <div style="font-family: monospace; color: #666; font-size: 13px;">{rec_hex.upper()}</div>
+    if not RECOMMENDATIONS_ENABLED:
+        rec_html = '<div style="color: #888; font-style: italic; padding: 20px; text-align: center;">🔍 Recommendations require training data (xkcd_scaled_data_Final.txt) — not available in this deployment.</div>'
+    else:
+        rec_html = '<div style="display: flex; flex-direction: column; gap: 8px;">'
+        for name, rec_hex, rec_rgb in recommendations:
+            rec_html += f""""
+            <div style="display: flex; align-items: center; padding: 10px; background: #f9f9f9; border-radius: 8px; border: 1px solid #eee;">
+                <div style="
+                    width: 50px; 
+                    height: 50px; 
+                    background-color: {rec_hex}; 
+                    border: 1px solid #ddd; 
+                    border-radius: 6px;
+                    margin-right: 16px;
+                    flex-shrink: 0;
+                "></div>
+                <div>
+                    <div style="font-weight: 600; font-size: 14px;">{name}</div>
+                    <div style="font-family: monospace; color: #666; font-size: 13px;">{rec_hex.upper()}</div>
+                </div>
             </div>
-        </div>
-        """
-    rec_html += '</div>'
+            """
+        rec_html += '</div>'
     
     # Color space conversions
     L, a, b_lab = rgb_to_lab(r, g, b)
@@ -363,6 +381,25 @@ with gr.Blocks(css=css, title="Color Name to RGB Predictor") as demo:
         inputs=color_input,
         label="Try these examples"
     )
+
+    # Citation
+    gr.Markdown("""
+    ---
+    ### 📚 Citation
+    If you use this work in your research, please cite:
+    ```bibtex
+    @article{jyothi2023text2color,
+      title={Text2Color Networks: Deep Learning Models for Color Generation from Compositional Color Descriptions},
+      author={Jyothi, Kondalarao and Okade, Manish},
+      journal={International Journal on Artificial Intelligence Tools},
+      volume={32},
+      number={06},
+      pages={2350026},
+      year={2023},
+      publisher={World Scientific}
+    }
+    ```
+    """)
     
     predict_btn.click(
         fn=process_color_name,