import os import requests from io import BytesIO from PIL import Image import pandas as pd import google.generativeai as genai import matplotlib.pyplot as plt from google.colab import files # Set up the Generative AI API key api_key = os.getenv('GOOGLE_API_KEY') # Use environment variable for API key security genai.configure(api_key=api_key) categories = ["Personal Care", "Household Care", "Dairy", "Staples", "Snacks and Beverages", "Packaged Food", "Fruits and Vegetables"] # Step 1: Download image from URL def download_image(image_url): try: response = requests.get(image_url) response.raise_for_status() # Check if the request was successful img = Image.open(BytesIO(response.content)) temp_path = "temp_image.jpg" # Temporary path img.save(temp_path) # Save the image locally for further use return temp_path except Exception as e: print(f"Error downloading image: {e}") return None # Step 2: Upload Image to the API def upload_image(image_path): sample_file = genai.upload_file(path=image_path, display_name="Product Image") print(f"Uploaded file '{sample_file.display_name}' as: {sample_file.uri}") return sample_file # Step 3: Display Image def display_image(image_path): img = Image.open(image_path) plt.imshow(img) plt.axis('off') plt.show() # Step 4: Classify image to decide whether it contains fruits/vegetables or other products def classify_image(sample_file): model = genai.GenerativeModel(model_name="gemini-1.5-pro-latest") response = model.generate_content([sample_file, "Does this image contain fruits or vegetables? Answer 'yes' or 'no' only."]) classification = response.text.strip().lower() return classification == "yes" # Step 5: Predict freshness (for fruits and vegetables) def predict_freshness(sample_file): model = genai.GenerativeModel(model_name="gemini-1.5-pro-latest") response = model.generate_content([sample_file, "Can you provide the average freshness index (1-10) of the fruits/vegetables in the image. Just output the number."]) try: freshness_index = int(response.text.strip()) return freshness_index except ValueError: print("Error: Unable to convert the response to an integer.") return None # Step 6: Generate product details (for other products) def generate_product_details(sample_file): model = genai.GenerativeModel(model_name="gemini-1.5-pro-latest") response = model.generate_content([sample_file, f"Tell me the name of each product, its category among the following list of categories: {categories}, brand, MRP, manufacturer, expiry date, and quantity in the image. " "Do not output anything else. Output format for each product: " "Product Name: [Extracted Product name], Category: [Extracted Category], Brand: [Extracted Brand name], MRP: [Extracted MRP], Manufacturer: [Extracted Manufacturer name], " "Expiry Date: [Extracted Expiry Date], Quantity: [Extracted Quantity]. Separate the details of each product with one newline character. " "If some of the information is not available for a product, then output NA for that detail." ]) return response.text.strip() if response else "" # Step 7: Parse the response into a DataFrame def parse_response_to_dataframe(response_text): columns = ["Product Name", "Category", "Brand", "MRP", "Manufacturer", "Expiry Date", "Quantity"] product_sections = response_text.split("\n") products_list = [] for product_section in product_sections: product_details = {col: "NA" for col in columns} response_parts = product_section.split(", ") for part in response_parts: if "Product Name" in part: product_details["Product Name"] = part.split(": ")[1] elif "Category" in part: product_details["Category"] = part.split(": ")[1] elif "Brand" in part: product_details["Brand"] = part.split(": ")[1] elif "MRP" in part: product_details["MRP"] = part.split(": ")[1] elif "Manufacturer" in part: product_details["Manufacturer"] = part.split(": ")[1] elif "Expiry Date" in part: product_details["Expiry Date"] = part.split(": ")[1] elif "Quantity" in part: product_details["Quantity"] = part.split(": ")[1] products_list.append(product_details) return pd.DataFrame(products_list, columns=columns) # Step 8: Style the DataFrame for better display def style_dataframe(df): return df.style.set_properties(**{'text-align': 'center', 'border': '1px solid grey'}) .set_table_styles([{'selector': 'td', 'props': [('border', '1px solid grey')]}], overwrite=False) # Step 9: Display results (image and styled DataFrame) def display_results(image_path, styled_df): display_image(image_path) # Display the image print("\nProduct Details:\n") display(styled_df) # Display the styled DataFrame # Step 10: Save DataFrame to CSV def save_dataframe_to_csv(df, file_name="product_details.csv"): df.to_csv(file_name, index=False) print(f"DataFrame saved to {file_name}") # Combined Pipeline: Choose action based on image content def combined_pipeline(image_source, is_url=False): # Step 1: Download the image if it's a URL if is_url: image_path = download_image(image_source) if not image_path: print("Failed to download the image.") return else: image_path = image_source # Step 2: Upload the image sample_file = upload_image(image_path) if not sample_file: print("Error uploading image.") return # Step 3: Classify whether the image contains fruits/vegetables is_fruits_or_vegetables = classify_image(sample_file) if is_fruits_or_vegetables: print("Image contains fruits or vegetables. Predicting freshness...") freshness_index = predict_freshness(sample_file) if freshness_index is not None: print(f"The predicted freshness index is: {freshness_index}") else: print("Failed to predict freshness.") else: print("Image contains products. Extracting details...") response_text = generate_product_details(sample_file) if not response_text: print("No product details generated.") return df = parse_response_to_dataframe(response_text) styled_df = style_dataframe(df) display_results(image_path, styled_df) # Save the DataFrame to a CSV file save_dataframe_to_csv(df, "product_details.csv") # Download the CSV file files.download("product_details.csv")