feat: add RandomCrop module in transforms (#448)

* feat: add sar ckpt + perf

* feat: add randomcrop

* fix: test

* fix: test

* fix: scale and ratio

* fix: unused file

* fix: unused file

* fix: typo

* fix: typo

* fix: requested changes

* fix: flake8

* fix: test

* feat: add entry in docs

docs/source/transforms.rst

@@ -22,6 +22,7 @@ Here are all transformations that are available through DocTR:
 .. autoclass:: RandomGamma
 .. autoclass:: RandomJpegQuality
 .. autoclass:: RandomRotate
+.. autoclass:: RandomCrop
 
 
 Composing transformations

doctr/transforms/modules/base.py

@@ -4,14 +4,15 @@
 # See LICENSE or go to <https://www.apache.org/licenses/LICENSE-2.0.txt> for full license details.
 
 import random
+import math
 from typing import List, Any, Callable, Dict, Tuple
 import numpy as np
 
 from doctr.utils.repr import NestedObject
 from .. import functional as F
 
 
-__all__ = ['ColorInversion', 'OneOf', 'RandomApply', 'RandomRotate']
+__all__ = ['ColorInversion', 'OneOf', 'RandomApply', 'RandomRotate', 'RandomCrop']
 
 
 class ColorInversion(NestedObject):
@@ -89,14 +90,14 @@ def __call__(self, img: Any) -> Any:
 
 
 class RandomRotate(NestedObject):
-    """Randomly rotate a tensor image
+    """Randomly rotate a tensor image and its boxes
 
     Args:
         max_angle: maximum angle for rotation, in degrees. Angles will be uniformly picked in
             [-max_angle, max_angle]
         expand: whether the image should be padded before the rotation
     """
-    def __init__(self, max_angle: float = 25., expand: bool = False) -> None:
+    def __init__(self, max_angle: float = 5., expand: bool = False) -> None:
         self.max_angle = max_angle
         self.expand = expand
 
@@ -105,5 +106,31 @@ def extra_repr(self) -> str:
 
     def __call__(self, img: Any, target: Dict[str, np.ndarray]) -> Tuple[Any, Dict[str, np.ndarray]]:
         angle = random.uniform(-self.max_angle, self.max_angle)
-        img, target['boxes'] = F.rotate(img, target['boxes'], angle, self.expand)
-        return img, target
+        r_img, r_boxes = F.rotate(img, target["boxes"], angle, self.expand)
+        return r_img, dict(boxes=r_boxes)
+
+
+class RandomCrop(NestedObject):
+    """Randomly crop a tensor image and its boxes
+
+    Args:
+        scale: tuple of floats, relative (min_area, max_area) of the crop
+        ratio: tuple of float, relative (min_ratio, max_ratio) where ratio = h/w
+    """
+    def __init__(self, scale: Tuple[float, float] = (0.08, 1.), ratio: Tuple[float, float] = (0.75, 1.33)) -> None:
+        self.scale = scale
+        self.ratio = ratio
+
+    def extra_repr(self) -> str:
+        return f"scale={self.scale}, ratio={self.ratio}"
+
+    def __call__(self, img: Any, target: Dict[str, np.ndarray]) -> Tuple[Any, Dict[str, np.ndarray]]:
+        h, w = img.shape[:2]
+        random_scale = random.uniform(self.scale[0], self.scale[1])
+        random_ratio = random.uniform(self.ratio[0], self.ratio[1])
+        crop_h = math.sqrt(random_scale * random_ratio)
+        crop_w = math.sqrt(random_scale / random_ratio)
+        start_x, start_y = random.uniform(0, 1 - crop_w), random.uniform(0, 1 - crop_h)
+        crop_box = (int(start_x * w), int(start_y * h), int((start_x + crop_w) * w), int((start_y + crop_h) * h))
+        croped_img, crop_boxes = F.crop_detection(img, target["boxes"], crop_box)
+        return croped_img, dict(boxes=crop_boxes)

test/pytorch/test_transforms_pt.py

@@ -3,7 +3,7 @@
 import math
 import torch
 import numpy as np
-from doctr.transforms import Resize, ColorInversion, RandomRotate
+from doctr.transforms import Resize, ColorInversion, RandomRotate, RandomCrop
 from doctr.transforms.functional import rotate, crop_detection
 
 
@@ -86,7 +86,7 @@ def test_rotate():
     boxes = np.array([
         [15, 20, 35, 30]
     ])
-    r_img, r_boxes = rotate(input_t, boxes, angle=12.)
+    r_img, r_boxes = rotate(input_t, boxes, angle=12., expand=False)
     assert r_img.shape == (3, 50, 50)
     assert r_img[0, 0, 0] == 0.
     assert r_boxes.all() == np.array([[25., 25., 20., 10., 12.]]).all()
@@ -110,14 +110,14 @@ def test_rotate():
 
 
 def test_random_rotate():
-    rotator = RandomRotate(max_angle=10.)
+    rotator = RandomRotate(max_angle=10., expand=False)
     input_t = torch.ones((3, 50, 50), dtype=torch.float32)
     boxes = np.array([
         [15, 20, 35, 30]
     ])
-    r_img, target = rotator(input_t, dict(boxes=boxes))
+    r_img, r_boxes = rotator(input_t, dict(boxes=boxes))
     assert r_img.shape == input_t.shape
-    assert abs(target["boxes"][-1, -1]) <= 10.
+    assert abs(r_boxes["boxes"][-1, -1]) <= 10.
 
     # FP16 (only on GPU)
     if torch.cuda.is_available():
@@ -148,3 +148,15 @@ def test_crop_detection():
     img = torch.ones((3, 50, 50), dtype=torch.float16)
     c_img, _ = crop_detection(img, abs_boxes, crop_box)
     assert c_img.dtype == torch.float16
+
+
+def test_random_crop():
+    cropper = RandomCrop()
+    input_t = torch.ones((50, 50, 3), dtype=torch.float32)
+    boxes = np.array([
+        [15, 20, 35, 30]
+    ])
+    c_img, _ = cropper(input_t, dict(boxes=boxes))
+    new_h, new_w = c_img.shape[:2]
+    assert new_h >= 3
+    assert new_w >= 3

test/tensorflow/test_transforms_tf.py

@@ -251,7 +251,7 @@ def test_rotate():
     boxes = np.array([
         [15, 20, 35, 30]
     ])
-    r_img, r_boxes = rotate(input_t, boxes, angle=12.)
+    r_img, r_boxes = rotate(input_t, boxes, angle=12., expand=False)
     assert r_img.shape == (50, 50, 3)
     assert r_img[0, 0, 0] == 0.
     assert r_boxes.all() == np.array([[25., 25., 20., 10., 12.]]).all()
@@ -274,14 +274,14 @@ def test_rotate():
 
 
 def test_random_rotate():
-    rotator = T.RandomRotate(max_angle=10.)
+    rotator = T.RandomRotate(max_angle=10., expand=False)
     input_t = tf.ones((50, 50, 3), dtype=tf.float32)
     boxes = np.array([
         [15, 20, 35, 30]
     ])
-    r_img, target = rotator(input_t, dict(boxes=boxes))
+    r_img, r_boxes = rotator(input_t, dict(boxes=boxes))
     assert r_img.shape == input_t.shape
-    assert abs(target["boxes"][-1, -1]) <= 10.
+    assert abs(r_boxes["boxes"][-1, -1]) <= 10.
 
     # FP16
     input_t = tf.ones((50, 50, 3), dtype=tf.float16)
@@ -311,3 +311,15 @@ def test_crop_detection():
     img = tf.ones((50, 50, 3), dtype=tf.float16)
     c_img, _ = crop_detection(img, rel_boxes, crop_box)
     assert c_img.dtype == tf.float16
+
+
+def test_random_crop():
+    cropper = T.RandomCrop()
+    input_t = tf.ones((50, 50, 3), dtype=tf.float32)
+    boxes = np.array([
+        [15, 20, 35, 30]
+    ])
+    c_img, _ = cropper(input_t, dict(boxes=boxes))
+    new_h, new_w = c_img.shape[:2]
+    assert new_h >= 3
+    assert new_w >= 3