Refactor Streamlit app for audio transcription and remove facial emotion recognition logic

requirements.txt

@@ -1,6 +1,12 @@
+--extra-index-url https://download.pytorch.org/whl/cpu
+
+# Core app library
 streamlit==1.35.0
-Pillow==10.3.0
-numpy==1.26.4
-deepface==0.0.94
-tensorflow-cpu==2.16.1
-tf-keras==2.16.0
+
+# AI libraries for this test
+openai-whisper==20231117
+torch==2.7.0
+torchaudio==2.7.0
+
+# Utility for handling audio files
+soundfile

src/streamlit_app.py

@@ -1,56 +1,60 @@
 import os
 import streamlit as st
-
-# Point the cache directory to the guaranteed writable /tmp folder
-os.environ['DEEPFACE_HOME'] = '/tmp/.deepface'
-
-from PIL import Image
-import numpy as np
-from deepface import DeepFace
+import whisper
+import tempfile
 import logging
-import cv2
+
+# --- Set cache directory for models ---
+# Point the cache to a writable directory in the cloud environment
+os.environ['HF_HOME'] = '/tmp/huggingface'
 
 # --- Page Configuration ---
 st.set_page_config(
-    page_title="FER Test",
-    page_icon="😀",
+    page_title="Audio Transcription Test",
+    page_icon="🎤",
     layout="centered"
 )
 
-st.title("Step 1: Facial Emotion Recognition (FER) Test")
-st.write("Upload an image with a face to test the DeepFace library.")
+st.title("Step 2: Audio Transcription (Whisper) Test")
+st.write("Upload a short audio file (.wav, .mp3, .m4a) to test the Whisper model.")
 
 # --- Logger Configuration ---
 logging.basicConfig(level=logging.INFO)
-logging.getLogger('deepface').setLevel(logging.ERROR)
+logging.getLogger('huggingface_hub').setLevel(logging.WARNING)
+
+# --- Model Loading ---
+@st.cache_resource
+def load_whisper_model():
+    with st.spinner("Loading Whisper model... This may take a moment."):
+        # Load the base model and specify a writable cache directory
+        model = whisper.load_model("base", download_root="/tmp/whisper_cache")
+    return model
+
+whisper_model = load_whisper_model()
 
 
 # --- UI and Processing Logic ---
-uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
+uploaded_file = st.file_uploader("Choose an audio file...", type=["wav", "mp3", "m4a"])
 
 if uploaded_file is not None:
-    pil_image = Image.open(uploaded_file)
-    numpy_image = np.array(pil_image)
-    image_bgr = cv2.cvtColor(numpy_image, cv2.COLOR_RGB2BGR)
+    # Save the uploaded file to a temporary location
+    with tempfile.NamedTemporaryFile(delete=False, suffix=os.path.splitext(uploaded_file.name)[1]) as tfile:
+        tfile.write(uploaded_file.read())
+        temp_audio_path = tfile.name
 
-    st.image(pil_image, caption="Image Uploaded", use_column_width=True)
+    st.audio(temp_audio_path)
 
-    with st.spinner("Analyzing image for emotion..."):
+    with st.spinner("Transcribing audio..."):
         try:
-            # This will now download its models to /tmp/.deepface
-            analysis = DeepFace.analyze(
-                img_path=image_bgr,
-                actions=['emotion'],
-                enforce_detection=False,
-                silent=True
-            )
-
-            if isinstance(analysis, list) and len(analysis) > 0:
-                dominant_emotion = analysis[0]['dominant_emotion']
-                st.success(f"Dominant Emotion Detected: **{dominant_emotion.capitalize()}**")
-                st.write(analysis[0]['emotion'])
-            else:
-                st.warning("No face detected in the image.")
+            # Transcribe the audio file using Whisper
+            result = whisper_model.transcribe(temp_audio_path, fp16=False)
+            transcribed_text = result.get("text", "No text transcribed.")
+
+            st.success("Transcription Complete!")
+            st.markdown(f"> {transcribed_text}")
 
         except Exception as e:
-            st.error(f"An error occurred during analysis: {e}")
+            st.error(f"An error occurred during transcription: {e}")
+        finally:
+            # Clean up the temporary file
+            os.unlink(temp_audio_path)

src/streamlit_app_imageFER.py

@@ -0,0 +1,56 @@
+import os
+import streamlit as st
+
+# Point the cache directory to the guaranteed writable /tmp folder
+os.environ['DEEPFACE_HOME'] = '/tmp/.deepface'
+
+from PIL import Image
+import numpy as np
+from deepface import DeepFace
+import logging
+import cv2
+
+# --- Page Configuration ---
+st.set_page_config(
+    page_title="FER Test",
+    page_icon="😀",
+    layout="centered"
+)
+
+st.title("Step 1: Facial Emotion Recognition (FER) Test")
+st.write("Upload an image with a face to test the DeepFace library.")
+
+# --- Logger Configuration ---
+logging.basicConfig(level=logging.INFO)
+logging.getLogger('deepface').setLevel(logging.ERROR)
+
+
+# --- UI and Processing Logic ---
+uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
+
+if uploaded_file is not None:
+    pil_image = Image.open(uploaded_file)
+    numpy_image = np.array(pil_image)
+    image_bgr = cv2.cvtColor(numpy_image, cv2.COLOR_RGB2BGR)
+
+    st.image(pil_image, caption="Image Uploaded", use_column_width=True)
+
+    with st.spinner("Analyzing image for emotion..."):
+        try:
+            # This will now download its models to /tmp/.deepface
+            analysis = DeepFace.analyze(
+                img_path=image_bgr,
+                actions=['emotion'],
+                enforce_detection=False,
+                silent=True
+            )
+
+            if isinstance(analysis, list) and len(analysis) > 0:
+                dominant_emotion = analysis[0]['dominant_emotion']
+                st.success(f"Dominant Emotion Detected: **{dominant_emotion.capitalize()}**")
+                st.write(analysis[0]['emotion'])
+            else:
+                st.warning("No face detected in the image.")
+
+        except Exception as e:
+            st.error(f"An error occurred during analysis: {e}")