Browse Source

add hubserving

will-jl944 4 years ago
parent
commit
838809a031
3 changed files with 5 additions and 11 deletions
  1. 1 8
      paddlex/cv/models/detector.py
  2. 1 1
      paddlex/cv/models/utils/visualize.py
  3. 3 2
      paddlex/deploy.py

+ 1 - 8
paddlex/cv/models/detector.py

@@ -560,17 +560,10 @@ class BaseDetector(BaseModel):
                     if label == -1:
                         continue
                     category = self.labels[int(label)]
-                    import pycocotools.mask as mask_util
-                    rle = mask_util.encode(
-                        np.array(
-                            mask[:, :, None], order="F", dtype="uint8"))[0]
-                    if six.PY3:
-                        if 'counts' in rle:
-                            rle['counts'] = rle['counts'].decode("utf8")
                     sg_res = {
                         'category_id': int(label),
                         'category': category,
-                        'mask': rle,
+                        'mask': mask.astype('uint8'),
                         'score': score
                     }
                     seg_res.append(sg_res)

+ 1 - 1
paddlex/cv/models/utils/visualize.py

@@ -211,7 +211,7 @@ def draw_bbox_mask(image, results, threshold=0.5, color_map=None):
 
         # draw mask
         if 'mask' in dt:
-            mask = mask_util.decode(dt['mask']) * 255
+            mask = dt['mask'] * 255
             image = image.astype('float32')
             alpha = .7
             w_ratio = .4

+ 3 - 2
paddlex/deploy.py

@@ -175,8 +175,6 @@ class Predictor(object):
             logging.error(
                 "Invalid model type {}.".format(self._model.model_type),
                 exit=True)
-        if len(preds) == 1:
-            preds = preds[0]
 
         return preds
 
@@ -255,6 +253,9 @@ class Predictor(object):
         self.timer.repeats = repeats
         self.timer.info(average=True)
 
+        if isinstance(img_file, (str, np.ndarray)):
+            results = results[0]
+
         return results
 
     def batch_predict(self, image_list, **params):