Update app.py

app.py

@@ -11,92 +11,76 @@ os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"
 import gradio as gr
 import tensorflow as tf
 import numpy as np
-import pickle
 from PIL import Image
+from tensorflow.keras.preprocessing.text import tokenizer_from_json
 
-# ================================
-# CONSTANTS (MUST MATCH TRAINING)
-# ================================
-IMG_SIZE = (128, 128)
-MAX_LEN = 50
-MAX_WORDS = 20000
-LABELS = ["Critical", "High", "Medium", "Low"]
+print("✅ Imports loaded")
 
 # ================================
-# LOAD TOKENIZER
+# LOAD FUSION MODEL (KERAS 3 SAFE)
 # ================================
-with open("tokenizer.pkl", "rb") as f:
-    tokenizer = pickle.load(f)
+MODEL_PATH = "fusion_ticket_model_final.keras"
 
-print("✅ Tokenizer loaded")
+fusion_model = tf.keras.models.load_model(
+    MODEL_PATH,
+    compile=False
+)
+
+print("✅ Fusion model loaded")
 
 # ================================
-# BUILD FUSION MODEL ARCHITECTURE
+# LOAD TOKENIZER (JSON, NOT PICKLE)
 # ================================
+with open("tokenizer.json", "r") as f:
+    tokenizer = tokenizer_from_json(f.read())
 
-# ---- Image encoder ----
-image_input = tf.keras.Input(shape=(128, 128, 3), name="image_input")
-
-x = tf.keras.layers.Conv2D(32, (3, 3), activation="relu")(image_input)
-x = tf.keras.layers.MaxPooling2D((2, 2))(x)
-x = tf.keras.layers.Conv2D(64, (3, 3), activation="relu")(x)
-x = tf.keras.layers.MaxPooling2D((2, 2))(x)
-x = tf.keras.layers.Conv2D(128, (3, 3), activation="relu")(x)
-x = tf.keras.layers.MaxPooling2D((2, 2))(x)
-x = tf.keras.layers.Flatten()(x)
-image_features = tf.keras.layers.Dense(128, activation="relu")(x)
-
-# ---- Text encoder ----
-text_input = tf.keras.Input(shape=(MAX_LEN,), name="text_input")
-
-y = tf.keras.layers.Embedding(MAX_WORDS, 128)(text_input)
-y = tf.keras.layers.LSTM(64)(y)
-text_features = tf.keras.layers.Dense(64, activation="relu")(y)
-
-# ---- Fusion head ----
-combined = tf.keras.layers.Concatenate()([image_features, text_features])
-z = tf.keras.layers.Dense(128, activation="relu")(combined)
-z = tf.keras.layers.Dense(64, activation="relu")(z)
-output = tf.keras.layers.Dense(4, activation="softmax")(z)
-
-fusion_model = tf.keras.Model(
-    inputs=[image_input, text_input],
-    outputs=output
-)
+print("✅ Tokenizer loaded")
 
 # ================================
-# LOAD WEIGHTS (THIS IS THE KEY)
+# CONSTANTS (MUST MATCH TRAINING)
 # ================================
-fusion_model.load_weights("fusion_weights.weights.h5")
-print("✅ Fusion weights loaded")
+IMG_SIZE = (128, 128)
+MAX_LEN = 50
+LABELS = ["Critical", "High", "Medium", "Low"]
 
 # ================================
-# PREPROCESSING
+# IMAGE PREPROCESSING
 # ================================
 def preprocess_image(image: Image.Image):
     image = image.convert("RGB")
     image = image.resize(IMG_SIZE)
     img = np.array(image, dtype=np.float32) / 255.0
-    return np.expand_dims(img, axis=0)
+    img = np.expand_dims(img, axis=0)
+    return img
 
+# ================================
+# TEXT PREPROCESSING
+# ================================
 def preprocess_text(text: str):
     if text is None:
         text = ""
     seq = tokenizer.texts_to_sequences([text])
-    return tf.keras.preprocessing.sequence.pad_sequences(
+    padded = tf.keras.preprocessing.sequence.pad_sequences(
         seq, maxlen=MAX_LEN
     )
+    return padded
 
 # ================================
 # PREDICTION FUNCTION
 # ================================
 def predict_ticket(image, text):
     if image is None:
-        return {"Critical": 0, "High": 0, "Medium": 0, "Low": 0}
+        return {
+            "Critical": 0.0,
+            "High": 0.0,
+            "Medium": 0.0,
+            "Low": 0.0,
+        }
 
     img = preprocess_image(image)
     txt = preprocess_text(text)
 
+    # IMPORTANT: input order MUST match training
     probs = fusion_model.predict([img, txt], verbose=0)[0]
 
     return {
@@ -113,10 +97,26 @@ interface = gr.Interface(
     fn=predict_ticket,
     inputs=[
         gr.Image(type="pil", label="📤 Upload Ticket Screenshot"),
-        gr.Textbox(lines=4, label="✍️ Ticket Description")
+        gr.Textbox(
+            lines=4,
+            placeholder="Describe the issue (optional but recommended)",
+            label="✍️ Ticket Description"
+        )
     ],
-    outputs=gr.Label(num_top_classes=4, label="🚨 Predicted Severity"),
-    title="🎫 Ticket Severity Classification",
+    outputs=gr.Label(
+        num_top_classes=4,
+        label="🚨 Predicted Ticket Severity"
+    ),
+    title="🎫 Ticket Severity Classification System",
+    description=(
+        "This system uses a **CNN + NLP Fusion Model** trained on real ticket screenshots "
+        "and descriptions to predict ticket urgency.\n\n"
+        "**Severity Levels:** Critical | High | Medium | Low"
+    ),
+    allow_flagging="never"
 )
 
+# ================================
+# LAUNCH (HF SPACES SAFE)
+# ================================
 interface.launch(server_name="0.0.0.0", server_port=7860)