address comments

official/vision/beta/projects/yolo/dataloaders/yolo_input.py

@@ -14,7 +14,6 @@
 
 """Detection Data parser and processing for YOLO."""
 import tensorflow as tf
-import numpy as np
 from official.vision.beta.projects.yolo.ops import preprocessing_ops
 from official.vision.beta.projects.yolo.ops import anchor
 from official.vision.beta.ops import preprocess_ops
@@ -230,7 +229,7 @@ class Parser(parser.Parser):
           self._aug_rand_perspective)
 
       # Clip and clean boxes.
-      boxes, inds = preprocessing_ops.apply_infos(
+      boxes, inds = preprocessing_ops.transform_and_clip_boxes(
           boxes,
           infos,
           affine=affine,
@@ -284,7 +283,7 @@ class Parser(parser.Parser):
 
     # Clip and clean boxes.
     image = image / 255.0
-    boxes, inds = preprocessing_ops.apply_infos(
+    boxes, inds = preprocessing_ops.transform_and_clip_boxes(
         boxes, infos, shuffle_boxes=False, area_thresh=0.0, augment=True)
     classes = tf.gather(classes, inds)
     info = infos[-1]

official/vision/beta/projects/yolo/ops/mosaic.py

@@ -40,6 +40,7 @@ class Mosaic:
                random_pad=False,
                random_flip=False, 
                area_thresh=0.1,
+               pad_value=preprocessing_ops.PAD_VALUE, 
                seed=None):
     """Initializes parameters for mosaic.
 
@@ -95,6 +96,7 @@ class Mosaic:
     self._aug_rand_angle = aug_rand_angle
     self._aug_rand_perspective = aug_rand_perspective
     self._random_flip = random_flip
+    self._pad_value = pad_value
 
     self._deterministic = seed != None
     self._seed = seed if seed is not None else random.randint(0, 2**30)
@@ -169,7 +171,7 @@ class Mosaic:
         seed=self._seed)
 
     # Clip and clean boxes.
-    boxes, inds = preprocessing_ops.apply_infos(
+    boxes, inds = preprocessing_ops.transform_and_clip_boxes(
         boxes,
         infos,
         area_thresh=self._area_thresh,
@@ -195,15 +197,15 @@ class Mosaic:
               -center[1], center[1], seed=self._seed))
 
       # clip the boxes to those with in the image
-      image = tfa.image.translate(
-          image, [cw, ch], fill_value=preprocessing_ops.get_pad_value())
+      image = tfa.image.translate(image, [cw, ch], fill_value=self._pad_value)
       boxes = box_ops.denormalize_boxes(boxes, shape[:2])
       boxes = boxes + tf.cast([ch, cw, ch, cw], boxes.dtype)
       boxes = box_ops.clip_boxes(boxes, shape[:2])
       inds = box_ops.get_non_empty_box_indices(boxes)
 
       boxes = box_ops.normalize_boxes(boxes, shape[:2])
-      boxes, classes, is_crowd, area = self._select_ind(inds, boxes, classes, is_crowd, area)
+      boxes, classes, is_crowd, area = self._select_ind(
+        inds, boxes, classes, is_crowd, area)
 
 
     # warp and scale the fully stitched sample 
@@ -220,7 +222,7 @@ class Mosaic:
     image = tf.image.resize(image, (height, width))
 
     # clip and clean boxes
-    boxes, inds = preprocessing_ops.apply_infos(
+    boxes, inds = preprocessing_ops.transform_and_clip_boxes(
         boxes, None, affine=affine, area_thresh=self._area_thresh, 
         seed=self._seed)
     classes, is_crowd, area = self._select_ind(inds, classes, is_crowd, area)

official/vision/beta/projects/yolo/ops/preprocessing_ops.py

 import tensorflow as tf
 import numpy as np
 import random
-import os
 
 import tensorflow_addons as tfa
 from official.vision.beta.ops import box_ops as bbox_ops
@@ -9,7 +8,6 @@ from official.vision.beta.ops import box_ops as bbox_ops
 PAD_VALUE = 114
 GLOBAL_SEED_SET = False
 
-
 def set_random_seeds(seed=0):
   """Sets all accessible global seeds to properly apply randomization.
 
@@ -24,18 +22,11 @@ def set_random_seeds(seed=0):
   """  
   if seed is not None:
     global GLOBAL_SEED_SET
-    os.environ['PYTHONHASHSEED'] = str(seed)
     random.seed(seed)
     GLOBAL_SEED_SET = True
   tf.random.set_seed(seed)
   np.random.seed(seed)
 
-
-def get_pad_value():
-  """Return the padding value."""
-  return PAD_VALUE
-
-
 def rand_uniform_strong(minval, maxval, dtype=tf.float32, seed=None, shape=[]):
   """A unified function for consistent random number generation. 
   
@@ -394,7 +385,8 @@ def resize_and_jitter_image(image,
 
     # Letter box the image.
     if letter_box == True or letter_box is None:
-      image_aspect_ratio, input_aspect_ratio = original_width / original_height, width / height
+      (image_aspect_ratio, 
+      input_aspect_ratio) = original_width / original_height, width / height
       distorted_aspect = image_aspect_ratio / input_aspect_ratio
 
       delta_h, delta_w = 0.0, 0.0
@@ -473,7 +465,7 @@ def resize_and_jitter_image(image,
     # Pad the image to desired size.
     image_ = tf.pad(
         cropped_image, [[pad[0], pad[2]], [pad[1], pad[3]], [0, 0]],
-        constant_values=get_pad_value())
+        constant_values=PAD_VALUE)
     pad_info = tf.stack([
         tf.cast(tf.shape(cropped_image)[:2], tf.float32),
         tf.cast(tf.shape(image_)[:2], dtype=tf.float32),
@@ -660,7 +652,7 @@ def affine_warp_image(image,
   image = tfa.image.transform(
       image,
       affine,
-      fill_value=get_pad_value(),
+      fill_value=PAD_VALUE,
       output_shape=desired_size,
       interpolation='bilinear')
 
@@ -826,7 +818,7 @@ def resize_and_crop_boxes(boxes, image_scale, output_size, offset, box_history):
   return clipped_boxes, box_history
 
 
-def apply_infos(boxes,
+def transform_and_clip_boxes(boxes,
                 infos,
                 affine=None,
                 shuffle_boxes=False,