Upload 3 files

Dockerfile

@@ -0,0 +1,32 @@
+# Use official lightweight Python image
+FROM python:3.11-slim
+
+# Set environment variables
+ENV PYTHONDONTWRITEBYTECODE=1
+ENV PYTHONUNBUFFERED=1
+ENV TF_CPP_MIN_LOG_LEVEL=2
+ENV TF_ENABLE_ONEDNN_OPTS=0
+
+# Create a working directory
+WORKDIR /app
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+# Install Python dependencies
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the model files directly into the image (Since it's only ~92MB!)
+COPY model/ ./model/
+
+# Copy your application code
+COPY app.py .
+
+# Expose the port FastAPI runs on
+EXPOSE 8000
+
+# Command to run the application using uvicorn (Simpler for HuggingFace)
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "8000"]

app.py

@@ -0,0 +1,178 @@
+import os
+import io
+import cv2
+import logging
+import numpy as np
+
+# ==========================================
+# 0. PRODUCTION LOGGING & ENVIRONMENT SETUP
+# ==========================================
+# Set TensorFlow variables BEFORE importing TF to suppress warnings
+os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
+os.environ['TF_ENABLE_ONEDNN_OPTS'] = '0'
+
+import tensorflow as tf
+from tensorflow import keras
+from fastapi import FastAPI, UploadFile, File, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+
+# Configure production logging
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
+)
+logger = logging.getLogger("AiroDx-API")
+
+# ==========================================
+# 1. CONFIGURATION & PATHS
+# ==========================================
+# Use Environment Variables for paths (Allows Docker to change them without changing code)
+MODEL_PATH = os.getenv("MODEL_PATH", "./model/remedis_multiview_final.keras")
+REMEDIS_BASE_PATH = os.getenv("REMEDIS_BASE_PATH", "./model/cxr-50x1-remedis-m")
+
+IMAGE_SIZE = (448, 448)
+TARGET_DISEASES = [
+    'Atelectasis', 'Cardiomegaly', 'Consolidation', 'Emphysema', 'Hernia', 
+    'Infiltrates', 'Mass', 'Nodule', 'Pleural Effusion', 'Pleural Thickening', 
+    'Pneumonia', 'Pneumothorax', 'Pulmonary Edema', 'Pulmonary Fibrosis'
+]
+
+# ==========================================
+# 2. REDEFINE CUSTOM LAYER
+# ==========================================
+@tf.keras.utils.register_keras_serializable()
+class REMEDISBackbone(keras.layers.Layer):
+    def __init__(self, model_path=None, **kwargs):
+        # Accept legacy serialized model_path, but force the local deployment path.
+        super().__init__(**kwargs)
+        self.model_path = REMEDIS_BASE_PATH
+        self._remedis_model = None
+        
+    def build(self, input_shape):
+        if self._remedis_model is None:
+            if not os.path.exists(self.model_path):
+                logger.error(f"Base REMEDIS model not found at {self.model_path}")
+                raise FileNotFoundError(f"Base REMEDIS model not found at {self.model_path}")
+            self._remedis_model = tf.saved_model.load(self.model_path)
+        super().build(input_shape)
+    
+    def call(self, inputs, training=None):
+        return self._remedis_model(inputs)
+    
+    def compute_output_shape(self, input_shape):
+        return (input_shape[0], 14, 14, 2048)
+    
+    def get_config(self):
+        config = super().get_config()
+        return config
+    
+    @classmethod
+    def from_config(cls, config):
+        # Remove stale serialized path (e.g., Kaggle) before layer construction.
+        config.pop("model_path", None)
+        return cls(**config)
+
+# ==========================================
+# 3. INITIALIZE APP & LOAD MODEL
+# ==========================================
+app = FastAPI(
+    title="AiroDx Multi-View Diagnostic API", 
+    description="Multimodal diagnostic support using Siamese REMEDIS network.",
+    version="1.0.0"
+)
+
+# CORS setup for production (Can be restricted via ENV variable)
+ALLOWED_ORIGINS = os.getenv("ALLOWED_ORIGINS", "*").split(",")
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=ALLOWED_ORIGINS, 
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+logger.info(f"Loading weights from {MODEL_PATH}...")
+try:
+    model = keras.models.load_model(
+        MODEL_PATH, 
+        custom_objects={"REMEDISBackbone": REMEDISBackbone}
+    )
+    logger.info("Model loaded successfully into memory!")
+except Exception as e:
+    logger.critical(f"Failed to load model: {e}")
+    model = None
+
+# ==========================================
+# 4. HEALTH CHECK ENDPOINT (For Cloud Hosting)
+# ==========================================
+@app.get("/health")
+async def health_check():
+    """Endpoint used by load balancers and Docker to ensure API is alive."""
+    if model is None:
+        raise HTTPException(status_code=503, detail="Model failed to load.")
+    return {"status": "healthy", "model_ready": True}
+
+# ==========================================
+# 5. PREPROCESSING FUNCTION
+# ==========================================
+async def process_image(file: UploadFile) -> np.ndarray:
+    try:
+        # Read the file bytes
+        contents = await file.read()
+        nparr = np.frombuffer(contents, np.uint8)
+        img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+        
+        if img is None:
+            raise ValueError(f"Invalid image file: {file.filename}")
+            
+        # Exact same preprocessing used in training
+        img = cv2.resize(img, IMAGE_SIZE)
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img = img.astype('float32') / 255.0
+        
+        # Add batch dimension: (448, 448, 3) -> (1, 448, 448, 3)
+        return np.expand_dims(img, axis=0)
+    except Exception as e:
+        logger.error(f"Image processing error: {str(e)}")
+        raise HTTPException(status_code=400, detail=f"Error processing image: {str(e)}")
+
+# ==========================================
+# 6. PREDICTION ENDPOINT
+# ==========================================
+@app.post("/predict")
+async def predict(
+    frontal_image: UploadFile = File(..., description="The PA/AP Frontal X-ray"),
+    lateral_image: UploadFile = File(..., description="The Lateral X-ray")
+):
+    logger.info(f"Received prediction request for files: {frontal_image.filename}, {lateral_image.filename}")
+    
+    # 1. Process both images
+    img_pa = await process_image(frontal_image)
+    img_l = await process_image(lateral_image)
+    
+    # 2. Run inference
+    predictions = model.predict({
+        "pa_image": img_pa,
+        "l_image": img_l
+    })
+    
+    # 3. Format results
+    probabilities = predictions[0].tolist() 
+    
+    results = []
+    for disease, prob in zip(TARGET_DISEASES, probabilities):
+        results.append({
+            "disease": disease,
+            "probability": round(prob, 4),
+            "flagged": bool(prob > 0.5) # You can adjust this threshold later based on your F1 tuning
+        })
+    
+    # Sort results by probability (highest first)
+    results.sort(key=lambda x: x["probability"], reverse=True)
+    
+    logger.info("Prediction processed successfully.")
+    
+    return {
+        "status": "success",
+        "predictions": results
+    }

requirements.txt

@@ -0,0 +1,44 @@
+absl-py==2.4.0
+annotated-doc==0.0.4
+annotated-types==0.7.0
+anyio==4.13.0
+astunparse==1.6.3
+certifi==2026.2.25
+charset-normalizer==3.4.6
+click==8.3.1
+colorama==0.4.6
+fastapi==0.135.2
+flatbuffers==25.12.19
+gast==0.7.0
+google-pasta==0.2.0
+grpcio==1.78.0
+h11==0.16.0
+h5py==3.14.0
+idna==3.11
+keras==3.13.2
+libclang==18.1.1
+markdown-it-py==4.0.0
+mdurl==0.1.2
+ml_dtypes==0.5.4
+namex==0.1.0
+numpy==2.4.3
+opencv-python-headless==4.13.0.92
+opt_einsum==3.4.0
+optree==0.19.0
+packaging==26.0
+protobuf==7.34.1
+pydantic==2.12.5
+pydantic_core==2.41.5
+Pygments==2.19.2
+python-multipart==0.0.22
+requests==2.33.0
+rich==14.3.3
+six==1.17.0
+starlette==1.0.0
+tensorflow==2.21.0
+termcolor==3.3.0
+typing-inspection==0.4.2
+typing_extensions==4.15.0
+urllib3==2.6.3
+uvicorn==0.42.0
+wrapt==2.1.2