Add roi perspective transform op. (#13176)

* Add roi perspective transform.

* Add roi_perspective_transform_op.

* Fix code style.

* Add python api and fix doc.

* Fix API.spec

* Fix python api.

* Fix API.spec

* Move src to detection.

paddle/fluid/API.spec

@@ -296,6 +296,7 @@ paddle.fluid.layers.ssd_loss ArgSpec(args=['location', 'confidence', 'gt_box', '
 paddle.fluid.layers.detection_map ArgSpec(args=['detect_res', 'label', 'class_num', 'background_label', 'overlap_threshold', 'evaluate_difficult', 'has_state', 'input_states', 'out_states', 'ap_version'], varargs=None, keywords=None, defaults=(0, 0.3, True, None, None, None, 'integral'))
 paddle.fluid.layers.rpn_target_assign ArgSpec(args=['bbox_pred', 'cls_logits', 'anchor_box', 'anchor_var', 'gt_boxes', 'is_crowd', 'im_info', 'rpn_batch_size_per_im', 'rpn_straddle_thresh', 'rpn_fg_fraction', 'rpn_positive_overlap', 'rpn_negative_overlap', 'use_random'], varargs=None, keywords=None, defaults=(256, 0.0, 0.5, 0.7, 0.3, True))
 paddle.fluid.layers.anchor_generator ArgSpec(args=['input', 'anchor_sizes', 'aspect_ratios', 'variance', 'stride', 'offset', 'name'], varargs=None, keywords=None, defaults=(None, None, [0.1, 0.1, 0.2, 0.2], None, 0.5, None))
+paddle.fluid.layers.roi_perspective_transform ArgSpec(args=['input', 'rois', 'transformed_height', 'transformed_width', 'spatial_scale'], varargs=None, keywords=None, defaults=(1.0,))
 paddle.fluid.layers.generate_proposal_labels ArgSpec(args=['rpn_rois', 'gt_classes', 'is_crowd', 'gt_boxes', 'im_info', 'batch_size_per_im', 'fg_fraction', 'fg_thresh', 'bg_thresh_hi', 'bg_thresh_lo', 'bbox_reg_weights', 'class_nums', 'use_random'], varargs=None, keywords=None, defaults=(256, 0.25, 0.25, 0.5, 0.0, [0.1, 0.1, 0.2, 0.2], None, True))
 paddle.fluid.layers.generate_proposals ArgSpec(args=['scores', 'bbox_deltas', 'im_info', 'anchors', 'variances', 'pre_nms_top_n', 'post_nms_top_n', 'nms_thresh', 'min_size', 'eta', 'name'], varargs=None, keywords=None, defaults=(6000, 1000, 0.5, 0.1, 1.0, None))
 paddle.fluid.layers.iou_similarity ArgSpec(args=[], varargs='args', keywords='kwargs', defaults=None)

paddle/fluid/operators/detection/CMakeLists.txt

@@ -31,5 +31,6 @@ polygon_box_transform_op.cu)
 detection_library(rpn_target_assign_op SRCS rpn_target_assign_op.cc)
 detection_library(generate_proposal_labels_op SRCS generate_proposal_labels_op.cc)
 detection_library(generate_proposals_op SRCS generate_proposals_op.cc)
+detection_library(roi_perspective_transform_op SRCS roi_perspective_transform_op.cc roi_perspective_transform_op.cu)
 #Export local libraries to parent
 set(DETECTION_LIBRARY ${LOCAL_DETECTION_LIBS} PARENT_SCOPE)

python/paddle/fluid/layers/detection.py

@@ -39,6 +39,7 @@ __all__ = [
     'detection_map',
     'rpn_target_assign',
     'anchor_generator',
+    'roi_perspective_transform',
     'generate_proposal_labels',
     'generate_proposals',
 ]
@@ -1262,6 +1263,54 @@ def anchor_generator(input,
     return anchor, var
 
 
+def roi_perspective_transform(input,
+                              rois,
+                              transformed_height,
+                              transformed_width,
+                              spatial_scale=1.0):
+    """
+    ROI perspective transform op.
+
+    Args:
+        input (Variable): The input of ROIPerspectiveTransformOp. The format of 
+                          input tensor is NCHW. Where N is batch size, C is the
+                          number of input channels, H is the height of the feature,
+                          and W is the width of the feature.
+        rois (Variable):  ROIs (Regions of Interest) to be transformed. It should be
+                          a 2-D LoDTensor of shape (num_rois, 8). Given as 
+                          [[x1, y1, x2, y2, x3, y3, x4, y4], ...], (x1, y1) is the 
+                          top left coordinates, and (x2, y2) is the top right 
+                          coordinates, and (x3, y3) is the bottom right coordinates, 
+                          and (x4, y4) is the bottom left coordinates.
+        transformed_height (integer): The height of transformed output.
+        transformed_height (integer): The width of transformed output.
+        spatial_scale (float): Spatial scale factor to scale ROI coords. Default: 1.0
+
+    Returns:
+        Variable: The output of ROIPerspectiveTransformOp which is a 4-D tensor with shape 
+                  (num_rois, channels, transformed_h, transformed_w).
+
+    Examples:
+        .. code-block:: python
+
+            out = fluid.layers.roi_perspective_transform(input, rois, 7, 7, 1.0)
+    """
+    helper = LayerHelper('roi_perspective_transform', **locals())
+    dtype = helper.input_dtype()
+    out = helper.create_tmp_variable(dtype)
+    helper.append_op(
+        type="roi_perspective_transform",
+        inputs={"X": input,
+                "ROIs": rois},
+        outputs={"Out": out},
+        attrs={
+            "transformed_height": transformed_height,
+            "transformed_width": transformed_width,
+            "spatial_scale": spatial_scale
+        })
+    return out
+
+
 def generate_proposal_labels(rpn_rois,
                              gt_classes,
                              is_crowd,

python/paddle/fluid/tests/unittests/test_layers.py

@@ -573,6 +573,16 @@ class TestBook(unittest.TestCase):
             self.assertIsNotNone(out)
         print(str(program))
 
+    def test_roi_perspective_transform(self):
+        program = Program()
+        with program_guard(program):
+            x = layers.data(name="x", shape=[256, 30, 30], dtype="float32")
+            rois = layers.data(
+                name="rois", shape=[8], dtype="float32", lod_level=1)
+            output = layers.roi_perspective_transform(x, rois, 7, 7, 0.6)
+            self.assertIsNotNone(output)
+        print(str(program))
+
     def test_sequence_enumerate(self):
         program = Program()
         with program_guard(program):