Upload 3 files

Dockerfile

@@ -0,0 +1,14 @@
+FROM python:3.9
+
+WORKDIR /code
+
+COPY ./requirements.txt /code/requirements.txt
+
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+
+COPY . /code
+
+# Create a directory for uploads if it doesn't exist
+RUN mkdir -p /code/uploads && chmod 777 /code/uploads
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

main.py

@@ -0,0 +1,110 @@
+from fastapi import FastAPI, UploadFile, File, HTTPException
+from fastapi.middleware.cors import CORSMiddleware
+from PIL import Image, ImageChops, ImageEnhance
+import io
+import numpy as np
+
+app = FastAPI()
+
+# Allow CORS for frontend
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],  # In production, specify the frontend domain
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+def analyze_image_heuristics(image: Image.Image) -> float:
+    """
+    Enhanced heuristic analysis for AI detection.
+    Combines:
+    1. Metadata Analysis (EXIF, Software tags)
+    2. Error Level Analysis (ELA)
+    3. Frequency Analysis (FFT)
+    """
+    score = 0.0
+    
+    # --- 1. Metadata Check ---
+    exif_data = image.getexif()
+    if not exif_data:
+        score += 0.3  # Suspicious: No EXIF
+    else:
+        # Check for common AI generation software tags
+        # 305 is the tag for 'Software'
+        software = exif_data.get(305, "").lower()
+        ai_signatures = ['stable diffusion', 'midjourney', 'dall-e', 'comfyui']
+        if any(sig in software for sig in ai_signatures):
+            return 1.0 # Definitely AI
+            
+    # --- 2. Error Level Analysis (ELA) ---
+    try:
+        if image.mode != 'RGB':
+            image = image.convert('RGB')
+            
+        buffer = io.BytesIO()
+        image.save(buffer, 'JPEG', quality=90)
+        buffer.seek(0)
+        resaved = Image.open(buffer)
+        
+        ela_image = ImageChops.difference(image, resaved)
+        extrema = ela_image.getextrema()
+        max_diff = max([ex[1] for ex in extrema])
+        
+        # AI images often have very smooth noise profiles
+        if max_diff < 15: 
+            score += 0.2
+            
+    except Exception as e:
+        print(f"ELA Error: {e}")
+
+    # --- 3. Frequency Analysis (FFT) ---
+    # AI images often lack high-frequency details compared to real photos
+    try:
+        img_gray = image.convert('L')
+        f = np.fft.fft2(img_gray)
+        fshift = np.fft.fftshift(f)
+        magnitude_spectrum = 20 * np.log(np.abs(fshift) + 1)
+        
+        # Calculate mean magnitude at high frequencies
+        rows, cols = img_gray.size
+        crow, ccol = rows//2 , cols//2
+        # Mask center (low frequencies)
+        mask_size = 30
+        magnitude_spectrum[crow-mask_size:crow+mask_size, ccol-mask_size:ccol+mask_size] = 0
+        
+        high_freq_mean = np.mean(magnitude_spectrum)
+        
+        # Heuristic threshold: Real photos usually have more high-freq noise
+        if high_freq_mean < 100:
+            score += 0.3
+            
+    except Exception as e:
+        print(f"FFT Error: {e}")
+
+    return min(score, 0.99)
+
+@app.post("/analyze")
+async def analyze_media(file: UploadFile = File(...)):
+    try:
+        contents = await file.read()
+        image = Image.open(io.BytesIO(contents))
+        
+        # Perform analysis
+        ai_probability = analyze_image_heuristics(image)
+        
+        # Threshold
+        is_ai = ai_probability > 0.5
+        
+        return {
+            "filename": file.filename,
+            "is_ai": is_ai,
+            "confidence": round(ai_probability * 100, 2),
+            "message": "AI Generated" if is_ai else "Real / Human Created"
+        }
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+@app.get("/")
+def read_root():
+    return {"status": "Server is running"}

requirements.txt

@@ -0,0 +1,5 @@
+fastapi
+uvicorn
+python-multipart
+pillow
+numpy