add some methods

mmcv/image/transforms/__init__.py

 from .colorspace import *
 from .geometry import *
+from .normalize import *
 from .resize import *

mmcv/image/transforms/geometry.py

+from __future__ import division
+
 import cv2
 import numpy as np
 
-__all__ = ['imrotate', 'imcrop', 'impad']
+__all__ = ['imflip', 'imrotate', 'imcrop', 'impad', 'impad_to_multiple']
+
+
+def imflip(img, direction='horizontal'):
+    """Flip an image horizontally or vertically.
+
+    Args:
+        img (ndarray): Image to be flipped.
+        direction (str): The flip direction, either "horizontal" or "vertical".
+
+    Returns:
+        ndarray: The flipped image.
+    """
+    assert direction in ['horizontal', 'vertical']
+    if direction == 'horizontal':
+        return np.flip(img, axis=1)
+    else:
+        return np.flip(img, axis=0)
 
 
 def imrotate(img,
@@ -13,7 +32,7 @@ def imrotate(img,
     """Rotate an image.
 
     Args:
-        img (ndarray or str): Image to be rotated.
+        img (ndarray): Image to be rotated.
         angle (float): Rotation angle in degrees, positive values mean
             clockwise rotation.
         center (tuple): Center of the rotation in the source image, by default
@@ -146,13 +165,13 @@ def imcrop(img, bboxes, scale_ratio=1.0, pad_fill=None):
         return patches
 
 
-def impad(img, shape, pad_val):
+def impad(img, shape, pad_val=0):
     """Pad an image to a certain shape.
 
     Args:
         img (ndarray): Image to be padded.
         shape (tuple): Expected padding shape.
-        pad_val (number or list): Values to be filled in padding areas.
+        pad_val (number or sequence): Values to be filled in padding areas.
 
     Returns:
         ndarray: The padded image.
@@ -168,3 +187,19 @@ def impad(img, shape, pad_val):
     pad[...] = pad_val
     pad[:img.shape[0], :img.shape[1], ...] = img
     return pad
+
+
+def impad_to_multiple(img, divisor, pad_val=0):
+    """Pad an image to ensure each edge to be multiple to some number.
+
+    Args:
+        img (ndarray): Image to be padded.
+        divisor (int): Padded image edges will be multiple to divisor.
+        pad_val (number or sequence): Same as :func:`impad`.
+
+    Returns:
+        ndarray: The padded image.
+    """
+    pad_h = int(np.ceil(img.shape[0] / divisor)) * divisor
+    pad_w = int(np.ceil(img.shape[1] / divisor)) * divisor
+    return impad(img, (pad_h, pad_w), pad_val)

mmcv/image/transforms/normalize.py

+import numpy as np
+
+from .colorspace import bgr2rgb, rgb2bgr
+
+__all__ = ['imnorm', 'imdenorm']
+
+
+def imnorm(img, mean, std, to_rgb=True):
+    img = img.astype(np.float32)
+    if to_rgb:
+        img = bgr2rgb(img)
+    return (img - mean) / std
+
+
+def imdenorm(img, mean, std, to_bgr=True):
+    img = (img * std) + mean
+    if to_bgr:
+        img = rgb2bgr(img)
+    return img

mmcv/torchpack/hooks/logger/pavi.py

@@ -145,3 +145,17 @@ class PaviLoggerHook(LoggerHook):
         log_outs.pop('time', None)
         log_outs.pop('data_time', None)
         self.pavi.log(runner.mode, runner.iter, log_outs)
+
+
+def pavi_hook_connect(runner, cfg_filename, cfg_text):
+    for hook in runner.hooks:
+        if isinstance(hook, PaviLoggerHook):
+            hook.connect(
+                runner.model_name,
+                runner.work_dir,
+                info={
+                    'session_file': cfg_filename,
+                    'session_text': cfg_text
+                },
+                logger=runner.logger)
+            break

mmcv/torchpack/utils.py

 import functools
 import logging
+import sys
 import time
 from getpass import getuser
 from socket import gethostname
@@ -45,7 +46,7 @@ def add_file_handler(logger, filename=None, mode='w', level=logging.INFO):
     return logger
 
 
-def obj_from_dict(info, module, default_args=None):
+def obj_from_dict(info, parrent=None, default_args=None):
     """Initialize an object from dict.
 
     The dict must contain the key "type", which indicates the object type, it
@@ -60,14 +61,17 @@ def obj_from_dict(info, module, default_args=None):
             object.
 
     Returns:
-
+        any type: Object built from the dict.
     """
     assert isinstance(info, dict) and 'type' in info
     assert isinstance(default_args, dict) or default_args is None
     args = info.copy()
     obj_type = args.pop('type')
     if mmcv.is_str(obj_type):
-        obj_type = getattr(module, obj_type)
+        if parrent is not None:
+            obj_type = getattr(parrent, obj_type)
+        else:
+            obj_type = sys.modules[obj_type]
     elif not isinstance(obj_type, type):
         raise TypeError('type must be a str or valid type, but got {}'.format(
             type(obj_type)))

tests/test_image.py

@@ -178,6 +178,36 @@ class TestImage(object):
         resized_img, scale = mmcv.limit_size(img, 200, True)
         assert resized_img.shape == (150, 200, 3) and scale == 0.5
 
+    def test_imflip(self):
+        # test color image
+        img = np.random.rand(80, 60, 3)
+        h, w, c = img.shape
+        hf_img = mmcv.imflip(img)
+        assert hf_img.shape == (h, w, c)
+        for i in range(h):
+            for j in range(w):
+                for k in range(c):
+                    assert hf_img[i, j, k] == img[i, w - 1 - j, k]
+        hf_img = mmcv.imflip(img, direction='vertical')
+        assert hf_img.shape == (h, w, c)
+        for i in range(h):
+            for j in range(w):
+                for k in range(c):
+                    assert hf_img[i, j, k] == self.img[h - 1 - i, j, k]
+        # test grayscale image
+        img = np.random.rand(80, 60)
+        h, w, c = img.shape
+        hf_img = mmcv.imflip(img)
+        assert hf_img.shape == (h, w, c)
+        for i in range(h):
+            for j in range(w):
+                assert hf_img[i, j, k] == img[i, w - 1 - j]
+        hf_img = mmcv.imflip(img, direction='vertical')
+        assert hf_img.shape == (h, w, c)
+        for i in range(h):
+            for j in range(w):
+                assert hf_img[i, j, k] == self.img[h - 1 - i, j]
+
     def test_imcrop(self):
         # yapf: disable
         bboxes = np.array([[100, 100, 199, 199],  # center
@@ -246,6 +276,17 @@ class TestImage(object):
         with pytest.raises(AssertionError):
             mmcv.impad(img, (5, 5), [0, 1])
 
+    def test_impad_to_multiple(self):
+        img = np.random.rand(11, 14, 3).astype(np.float32)
+        padded_img = mmcv.impad_to_multiple(img, 4)
+        assert padded_img.shape == (12, 16, 3)
+        img = np.random.rand(20, 12).astype(np.float32)
+        padded_img = mmcv.impad_to_multiple(img, 5)
+        assert padded_img.shape == (20, 15)
+        img = np.random.rand(20, 12).astype(np.float32)
+        padded_img = mmcv.impad_to_multiple(img, 2)
+        assert padded_img.shape == (20, 12)
+
     def test_imrotate(self):
         img = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]).astype(np.uint8)
         assert_array_equal(mmcv.imrotate(img, 0), img)