Jas

adversarially_trained_model.keras

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1f376abeea93d25d15c2ae923320f11171f4b00c1b83b1ce4f9d910a6213fc30
-size 2738509
+oid sha256:dc920a45ed6a7cc03343f678c72605e979b7e4d25eee8de114a1d760eba74b53
+size 2781926

app.py

@@ -1,15 +1,16 @@
 import gradio as gr
 import tensorflow as tf
 import numpy as np
-import pandas as pd
 
 # 1. Load your trained .keras model
 # This line loads your entire model from the .keras file.
 model = tf.keras.models.load_model("adversarially_trained_model.keras")
 print("output_shape:", model.output_shape)           # e.g., (None, 5)
-CLASS_NAMES = ["normal", "dos", "probe", "r2l", "u2r"]
+CLASS_NAMES = [f'class_{i}' for i in range(23)]# Class names based on the attack mapping from the notebook
+print(CLASS_NAMES)
 
 def preprocess_single_record(record: dict, scaler, df_reference: pd.DataFrame) -> np.ndarray:
+    import pandas as pd
     
     # Convert input dict to DataFrame
     df_input = pd.DataFrame([record])