issue: shape of 4 models

app.py

@@ -2,6 +2,17 @@
 from fastapi import FastAPI
 
 app = FastAPI()
+from fastapi.middleware.cors import CORSMiddleware
+origins = [
+"*"
+]
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
 
 @app.get("/")
 def greet_json():
@@ -63,7 +74,6 @@ else:
     print(f"The file '{output_file}' does not exist at '{file_path}'.")
 
 #--------------------------------------------------------------------------------------------------------------------
-
 import os
 import numpy as np
 import tensorflow as tf
@@ -74,7 +84,10 @@ import matplotlib.pyplot as plt
 
 import os
 os.environ["TORCH_HOME"] = "/tmp/torch_cache"
-
+os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0"
+os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib_config"
+os.environ["FONTCONFIG_PATH"] = "/tmp/fontconfig"
+os.environ["HF_HOME"] = "/tmp/huggingface_cache"
 
 from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input
 from tensorflow.keras.models import Model
@@ -86,8 +99,8 @@ class UnifiedDeepfakeDetector:
     def __init__(self):
         self.input_shape = (224, 224, 3)
         self.vgg_model = self.build_vgg16_model()
-        self.dense_model = tf.keras.models.load_model('./downloads/file.h5')
-        self.cnn_model = tf.keras.models.load_model('./downloads/file2.h5')
+        self.dense_model = tf.keras.models.load_model('downloads/file2.h5')
+        self.cnn_model = tf.keras.models.load_model('downloads/file.h5')
         self.melody_machine = pipeline(model="MelodyMachine/Deepfake-audio-detection-V2")
 
     def build_vgg16_model(self):
@@ -178,7 +191,7 @@ class UnifiedDeepfakeDetector:
             # CNN Model Analysis
             if model_choice in ["CNN", "all"]:
                 mfcc = librosa.feature.mfcc(y=audio, sr=sr, n_mfcc=40)
-                mfcc_scaled = np.mean(mfcc.T, axis=0).reshape(1, 40, 1, 1)
+                mfcc_scaled = np.mean(mfcc.T, axis=0).reshape(None, 40, 1, 1)
                 pred = self.cnn_model.predict(mfcc_scaled)
                 results["CNN"] = {
                     "prediction": "FAKE" if np.argmax(pred[0]) == 0 else "REAL",
@@ -280,6 +293,7 @@ import librosa
 os.environ["TF_CPP_MIN_LOG_LEVEL"] = "2"
 os.environ["MPLCONFIGDIR"] = "/tmp/matplotlib"
 os.environ["FONTCONFIG_PATH"] = "/tmp/fontconfig"
+os.environ["TF_ENABLE_ONEDNN_OPTS"]="0"
 os.environ["HF_HOME"] = "/tmp/huggingface_cache"
 
 os.makedirs("/tmp/matplotlib", exist_ok=True)
@@ -299,6 +313,7 @@ def reencode_audio(input_path, output_path):
 #--------------------------------------------------------------------------------------------------------------------
 from collections import Counter
 from datetime import datetime
+import base64
 
 @app.post("/upload")
 async def upload_file(file: UploadFile = File(...)):
@@ -368,7 +383,7 @@ async def upload_file(file: UploadFile = File(...)):
     result = list(result)
     result.append("FAKE" if float(entropy) < 150 else "REAL")
     print(result)
-    r_normalized = [x.upper() for x in result]
+    r_normalized = [x.upper() for x in result if x is not None]
     counter = Counter(r_normalized)
 
     most_common_element, _ = counter.most_common(1)[0]

downloads/file.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1894bc00368d6b754f9ccbd7dec92c0fcd0fe741d23e87a92c3d54a3cc211b25
+size 4003296

downloads/file2.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3036383c72926c51fb8814b33f9bfb77f4f3d447fb34719c43319a252a4b741
+size 900560