import os import io from pathlib import Path import numpy as np import pandas as pd from PIL import Image import onnxruntime as ort from scipy.special import softmax import google.generativeai as genai from dotenv import load_dotenv # ===================== 0. LOAD ENV ===================== load_dotenv() GEMINI_API_KEY = os.getenv("GEMINI_API_KEY") if GEMINI_API_KEY: genai.configure(api_key=GEMINI_API_KEY) else: print("⚠️ WARNING: GEMINI_API_KEY not found. Image generation will be disabled.") # ===================== 1. PATHS ===================== # Ensure paths point to the assets folder (used by download_files.py) BASE_DIR = Path(__file__).resolve().parent / "assets" ONNX_MODEL_PATH = BASE_DIR / "ViT_PreProcessing-ops11-preprocessing-int-dynam_graph.onnx" TRAIN_METADATA_PATH = BASE_DIR / "train_metadata.csv" VENOM_XLSX_PATH = BASE_DIR / "venomstatus_with_antivenom.xlsx" COUNTRY_METADATA_PATH = BASE_DIR / "min-train_metadata.csv" # ===================== 2. SAFE MODEL LOADING ===================== session = None def load_model(): global session if session is None: if not ONNX_MODEL_PATH.exists(): raise FileNotFoundError(f"❌ ONNX model not found at {ONNX_MODEL_PATH}") session = ort.InferenceSession(str(ONNX_MODEL_PATH), providers=["CPUExecutionProvider"]) print("✅ Model loaded successfully!") return session # ===================== 3. IMAGE PREPROCESSING ===================== def preprocess_image(image: Image.Image) -> np.ndarray: image = image.convert("RGB").resize((224, 224)) arr = np.array(image, dtype=np.float32) / 255.0 arr = np.transpose(arr, (2, 0, 1)) return np.expand_dims(arr, axis=0) # ===================== 4. PREDICTION ===================== def predict_snake(image: Image.Image) -> dict: try: # Ensure all required files exist required_files = [TRAIN_METADATA_PATH, VENOM_XLSX_PATH, COUNTRY_METADATA_PATH] for path in required_files: if not path.exists(): raise FileNotFoundError(f"Missing file: {path}") metadata = pd.read_csv(TRAIN_METADATA_PATH) venom_data = pd.read_excel(VENOM_XLSX_PATH) country_metadata = pd.read_csv(COUNTRY_METADATA_PATH) class_names = metadata['binomial'].unique().tolist() input_tensor = preprocess_image(image) sess = load_model() input_name = sess.get_inputs()[0].name output_name = sess.get_outputs()[0].name logits = sess.run([output_name], {input_name: input_tensor})[0][0] probabilities = softmax(logits) predicted_class_id = int(np.argmax(probabilities)) species_name = class_names[predicted_class_id] confidence = float(np.max(probabilities)) # Lookup venom + antivenom data species_rows = venom_data[venom_data['species'] == species_name] species_row = species_rows.iloc[0].to_dict() if not species_rows.empty else {} # Country info country_list = country_metadata[country_metadata['binomial'] == species_name]['country'].unique() countries = [c for c in country_list if c.lower() not in ('unknown', 'unidentified', 'n/a')] country_output = ", ".join(countries) if countries else "Distribution data unavailable" return { "species_name": species_name, "venom_status": int(species_row.get('venom_status', -1)), "antivenom_name": species_row.get('antivenom Name', 'N/A'), "manufacturer": species_row.get('manufacturer', 'N/A'), "confidence": round(confidence, 4), "countries": country_output } except Exception as e: return {"error": "Prediction failed", "detail": str(e)} # ===================== 5. GEMINI IMAGE GENERATION ===================== def generate_image_from_text(prompt: str): if not GEMINI_API_KEY: raise RuntimeError("❌ Gemini API key not configured in environment.") full_prompt = f"Highly detailed, scientifically accurate photo of a snake: {prompt}" model = genai.GenerativeModel("imagen-3.0-generate-002") result = model.generate_content( prompt=full_prompt, generation_config={ "number_of_images": 1, "output_mime_type": "image/png", "aspect_ratio": "1:1" } ) if not result.generated_images: raise RuntimeError("Gemini API returned no image.") image_bytes = result.generated_images[0].image.image_bytes gen_image_pil = Image.open(io.BytesIO(image_bytes)).convert("RGB").resize((224, 224)) return image_bytes, gen_image_pil