fixed color pallete

app.py

@@ -30,14 +30,16 @@ import os
 os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'
 os.environ['OMP_NUM_THREADS'] = '4'
 
-app_version = 'dsdg_vid_2'
+app_version = 'dsdg_vid_3'
 
 device = torch.device("cpu")
 labels = ['Live', 'Spoof']
 PIX_THRESHOLD = 0.45
-DSDG_THRESHOLD = 50.0
+DSDG_THRESHOLD = 80.0
 DSDG_FACTOR = 1000000
+DSDG_PERCENTILE = 40
 MIN_FACE_WIDTH_THRESHOLD = 210
+
 examples = [
     ['examples/1_1_21_2_33_scene_fake.jpg'],
     ['examples/frame150_real.jpg'],
@@ -50,9 +52,9 @@ tfms = transforms.Compose([
     transforms.ToTensor(),
     transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
 ])
-deepix_model = DeePixBiS(pretrained=False)
-deepix_model.load_state_dict(torch.load('./DeePixBiS/DeePixBiS.pth'))
-deepix_model.eval()
+# deepix_model = DeePixBiS(pretrained=False)
+# deepix_model.load_state_dict(torch.load('./DeePixBiS/DeePixBiS.pth'))
+# deepix_model.eval()
 
 
 depth_config_path = 'tddfa/configs/mb1_120x120.yml'  # 'tddfa/configs/mb1_120x120.yml
@@ -166,7 +168,7 @@ def prepare_data_dsdg(images, boxes, depths):
         image_x[i, :, :, :] = cv.resize(image, (256, 256))
         # transform to binary mask --> threshold = 0 
         depth_x[i, :, :] = cv.resize(depth_img, (32, 32))
-    image_x = image_x[:, :, :, ::-1].transpose((0, 3, 1, 2))
+    image_x = image_x.transpose((0, 3, 1, 2))
     image_x = transform(image_x)
     image_x = torch.from_numpy(image_x.astype(float)).float()
     depth_x = torch.from_numpy(depth_x.astype(float)).float()
@@ -189,15 +191,12 @@ def dsdg_model_inference(imgs, bboxes, depth_imgs):
         for frame_t in range(inputs.shape[1]):
             mu, logvar, map_x, x_concat, x_Block1, x_Block2, x_Block3, x_input = cdcn_model(inputs[:, frame_t, :, :, :])
             score_norm = torch.sum(mu) / torch.sum(test_maps[:, frame_t, :, :])
-            scores.append(score_norm.item() * DSDG_FACTOR)
+            score = score_norm.item()
+            if score > 10:
+                score = 0.0
+            scores.append(score * DSDG_FACTOR)
             map_score += score_norm
-        map_score = map_score / inputs.shape[1]
-        map_score_list.append(map_score)
-    res_dsdg = map_score_list[0].item()
-    if res_dsdg > 10:
-        res_dsdg = 0.0
-    res_dsdg = res_dsdg * DSDG_FACTOR
-    return res_dsdg, scores
+    return scores
 
 
 def inference(img, dsdg_thresh):
@@ -244,7 +243,8 @@ def process_video(vid_path, dsdg_thresh):
     if not inference_images:
         return vid_path, {'Not supported right now': 0}, -1, vid_path, 'Faces too small or not found', -1
 
-    res_dsdg, scores = dsdg_model_inference(inference_images, inference_bboxes, inference_depths)
+    scores = dsdg_model_inference(inference_images, inference_bboxes, inference_depths)
+    res_dsdg = np.percentile(scores, DSDG_PERCENTILE)
     cls_dsdg = 'Real' if res_dsdg >= dsdg_thresh else 'Spoof'
     for img, bbox, score in zip(inference_images, inference_bboxes, scores):
         x, y, x2, y2 = bbox