Update app.py

app.py

@@ -35,7 +35,6 @@ import spaces
 plt.set_loglevel("ERROR")
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
-# Deep-fake specialist
 _face_det = MTCNN(select_largest=False, post_process=False, device=device).eval().to(device)
 _df_model = InceptionResnetV1(pretrained="vggface2", classify=True, num_classes=1, device=device)
 _df_model.load_state_dict(torch.load("resnet_inception.pth", map_location="cpu")["model_state_dict"])
@@ -88,7 +87,6 @@ def _extract_landmarks(rgb: np.ndarray) -> Tuple[np.ndarray, np.ndarray | None]:
 def _overlay_cam(cam, base):
     if torch.is_tensor(cam):  
         cam = cam.detach().cpu().numpy()
-    # ------------------------------------------------------
 
     cam = (cam - cam.min()) / (cam.max() - cam.min() + 1e-6)
     heat = Image.fromarray(
@@ -113,13 +111,13 @@ def _render_pdf(title: str, verdict: str, conf: dict, pages: List[Image.Image])
         pdf.cell(0, 6, f"Figure {idx+1}", ln=True)
 
         tmp = Path(tempfile.mktemp(suffix=".jpg"))
-        img.convert("RGB").save(tmp, format="JPEG")   # ← add .convert("RGB")
+        img.convert("RGB").save(tmp, format="JPEG")
         pdf.image(str(tmp), x=10, w=90)
         tmp.unlink(missing_ok=True)
     pdf.output(out)
     return str(out)
 
-# ────────────────────────── SuSy helpers (saliency) ────────────────────
+
 def _susy_cam(tensor: torch.Tensor, class_idx: int) -> np.ndarray:
     sal = Saliency(_susy_mod)
     grad = sal.attribute(tensor, target=class_idx).abs().mean(1, keepdim=True)
@@ -146,7 +144,6 @@ def _susy_predict(img: Image.Image):
         probs = _susy_mod(tens.to(device)).softmax(-1).mean(0).cpu().numpy()[1:]
     return dict(zip(_GEN_CLASSES, probs))
 
-# ───────────────────────────── fusion math ─────────────────────────────
 def _fuse(p_ai: float, p_df: float) -> float:
     return 1 - (1 - p_ai) * (1 - p_df)
 
@@ -201,7 +198,7 @@ def _predict_image(pil: Image.Image):
 
     return verdict, conf, gallery, bar_plot, pdf
 
-def _sample_idx(n):  # max 20 evenly spaced
+def _sample_idx(n):  
     return list(range(n)) if n <= MAX_SAMPLES else np.linspace(0, n-1, MAX_SAMPLES, dtype=int)
 
 @spaces.GPU
@@ -232,7 +229,8 @@ with gr.Blocks(css=_css, title="AI-Fake & Deepfake Analyser") as demo:
     gr.Markdown("""
     ## Deepfake detector
     Upload an **image** or a short **video**.  
-    The app fuses two complementary models, then shows heat-maps & a PDF report.
+    The app fuses two complementary models, then shows heat-maps & a PDF report.Made by Code Alchemists
+    Which is (Brijesh Khanoolkar, Shreeya Dessai, Slevin Rodrigues , Rafan Khan)
     """)
 
     with gr.Tab("Image"):