Spaces:
Running
Running
| import os | |
| import requests | |
| import openai | |
| import streamlit as st | |
| from PIL import Image | |
| from dotenv import load_dotenv | |
| import torch | |
| from transformers import AutoProcessor, AutoModelForImageClassification | |
| # Load environment variables from .env file | |
| load_dotenv() | |
| # Set up OpenAI API key | |
| openai.api_key = os.getenv('OPENAI_API_KEY') | |
| # Load the processor and model for facial expression recognition | |
| processor = AutoProcessor.from_pretrained("trpakov/vit-face-expression") | |
| model = AutoModelForImageClassification.from_pretrained("trpakov/vit-face-expression") | |
| # Function to query the facial expression recognition model | |
| def query_emotion(image): | |
| # Preprocess the image | |
| inputs = processor(images=image, return_tensors="pt") | |
| # Perform inference | |
| with torch.no_grad(): | |
| outputs = model(**inputs) | |
| # Get predicted class index (the class with the highest logit) | |
| logits = outputs.logits | |
| predicted_class_idx = torch.argmax(logits, dim=-1).item() | |
| # Retrieve the label names from the model | |
| label_names = model.config.id2label # Mapping of indices to emotion labels | |
| predicted_label = label_names[predicted_class_idx] # Get the predicted label | |
| return predicted_label | |
| # Function to generate a response using OpenAI based on detected emotion | |
| def generate_text_based_on_mood(emotion): | |
| try: | |
| prompt = f"Generate a light-hearted joke or motivational message for someone who is feeling {emotion}." | |
| # Call OpenAI's API using GPT-4 | |
| response = openai.ChatCompletion.create( | |
| model="gpt-4", # Specify the GPT-4 model | |
| messages=[ | |
| {"role": "user", "content": prompt} | |
| ] | |
| ) | |
| # Extract the generated text | |
| generated_text = response['choices'][0]['message']['content'] | |
| return generated_text.strip() | |
| except Exception as e: | |
| st.error(f"Error generating text: {e}") | |
| return "Sorry, I couldn't come up with a message at this moment." | |
| # Function to convert text to speech using gTTS | |
| def text_to_speech(text): | |
| from gtts import gTTS | |
| try: | |
| tts = gTTS(text, lang='en') | |
| audio_file = "output.mp3" | |
| tts.save(audio_file) # Save the audio file | |
| return audio_file | |
| except Exception as e: | |
| st.error(f"Error with TTS: {e}") | |
| return None | |
| # Streamlit UI | |
| st.title("Facial Expression Mood Detector") | |
| st.write("Upload an image of a face to detect mood and receive uplifting messages or jokes.") | |
| # Upload image | |
| uploaded_file = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"]) | |
| if uploaded_file is not None: | |
| # Load and display the image | |
| image = Image.open(uploaded_file) | |
| st.image(image, caption='Uploaded Image', use_column_width=True) | |
| # Detect facial expression | |
| emotion = query_emotion(image) | |
| st.write(f"Detected emotion: {emotion}") | |
| # Generate text based on detected emotion | |
| message = generate_text_based_on_mood(emotion) | |
| st.write("Here's something to cheer you up:") | |
| st.write(message) | |
| # Convert the generated message to audio | |
| audio_file = text_to_speech(message) | |
| # Provide an audio player in the Streamlit app if audio file exists | |
| if audio_file: | |
| st.audio(audio_file) # Streamlit will handle playback | |