Add gradio

musetalk/utils/preprocessing.py

@@ -40,6 +40,47 @@ def read_imgs(img_list):
         frames.append(frame)
     return frames
 
+def get_bbox_range(img_list,upperbondrange =0):
+    frames = read_imgs(img_list)
+    batch_size_fa = 1
+    batches = [frames[i:i + batch_size_fa] for i in range(0, len(frames), batch_size_fa)]
+    coords_list = []
+    landmarks = []
+    if upperbondrange != 0:
+        print('get key_landmark and face bounding boxes with the bbox_shift:',upperbondrange)
+    else:
+        print('get key_landmark and face bounding boxes with the default value')
+    average_range_minus = []
+    average_range_plus = []
+    for fb in tqdm(batches):
+        results = inference_topdown(model, np.asarray(fb)[0])
+        results = merge_data_samples(results)
+        keypoints = results.pred_instances.keypoints
+        face_land_mark= keypoints[0][23:91]
+        face_land_mark = face_land_mark.astype(np.int32)
+        
+        # get bounding boxes by face detetion
+        bbox = fa.get_detections_for_batch(np.asarray(fb))
+        
+        # adjust the bounding box refer to landmark
+        # Add the bounding box to a tuple and append it to the coordinates list
+        for j, f in enumerate(bbox):
+            if f is None: # no face in the image
+                coords_list += [coord_placeholder]
+                continue
+            
+            half_face_coord =  face_land_mark[29]#np.mean([face_land_mark[28], face_land_mark[29]], axis=0)
+            range_minus = (face_land_mark[30]- face_land_mark[29])[1]
+            range_plus = (face_land_mark[29]- face_land_mark[28])[1]
+            average_range_minus.append(range_minus)
+            average_range_plus.append(range_plus)
+            if upperbondrange != 0:
+                half_face_coord[1] = upperbondrange+half_face_coord[1] #手动调整  + 向下（偏29）  - 向上（偏28）
+
+    text_range=f"Total frame:「{len(frames)}」 Manually adjust range : [ -{int(sum(average_range_minus) / len(average_range_minus))}~{int(sum(average_range_plus) / len(average_range_plus))} ] , the current value: {upperbondrange}"
+    return text_range
+    
+
 def get_landmark_and_bbox(img_list,upperbondrange =0):
     frames = read_imgs(img_list)
     batch_size_fa = 1