adapt new im_shape im Mask R-CNN-FPN (#1760)

* adapt new im_shape im Mask R-CNN-FPN

* fix training

configs/_base_/readers/mask_reader.yml

@@ -17,14 +17,14 @@ TrainReader:
 
 EvalReader:
   inputs_def:
-    fields: ['image', 'im_info', 'im_id']
+    fields: ['image', 'im_shape', 'scale_factor', 'im_id']
   sample_transforms:
-  - DecodeImage: {to_rgb: true}
-  - NormalizeImage: {is_channel_first: false, is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
-  - ResizeImage: {interp: 1, max_size: 1333, target_size: 800, use_cv2: true}
-  - Permute: {channel_first: true, to_bgr: false}
+  - DecodeOp: {}
+  - NormalizeImageOp: {is_scale: true, mean: [0.485,0.456,0.406], std: [0.229, 0.224,0.225]}
+  - ResizeOp: {interp: 1, target_size: [800, 1333]}
+  - PermuteOp: {}
   batch_transforms:
-  - PadBatch: {pad_to_stride: 32, use_padded_im_info: false, pad_gt: false}
+  - PadBatchOp: {pad_to_stride: 32, pad_gt: false}
   batch_size: 1
   shuffle: false
   drop_last: false

ppdet/data/transform/operator.py

@@ -141,7 +141,7 @@ class DecodeOp(BaseOperator):
                 "image width.".format(im.shape[1], sample['w']))
             sample['w'] = im.shape[1]
 
-        sample['im_shape'] = np.array(im.shape[:2], dtype=np.int32)
+        sample['im_shape'] = np.array(im.shape[:2], dtype=np.float32)
         sample['scale_factor'] = np.array([1., 1.], dtype=np.float32)
         return sample
 
@@ -666,8 +666,8 @@ class ResizeOp(BaseOperator):
             im_scale = min(target_size_min / im_size_min,
                            target_size_max / im_size_max)
 
-            resize_h = int(im_scale * im_shape[0])
-            resize_w = int(im_scale * im_shape[1])
+            resize_h = im_scale * float(im_shape[0])
+            resize_w = im_scale * float(im_shape[1])
 
             im_scale_x = im_scale
             im_scale_y = im_scale
@@ -678,14 +678,14 @@ class ResizeOp(BaseOperator):
 
         im = self.apply_image(sample['image'], [im_scale_x, im_scale_y])
         sample['image'] = im
-        sample['im_shape'] = np.array([resize_h, resize_w], dtype=np.int32)
+        sample['im_shape'] = np.asarray([resize_h, resize_w], dtype=np.float32)
         if 'scale_factor' in sample:
             scale_factor = sample['scale_factor']
-            sample['scale_factor'] = np.array(
+            sample['scale_factor'] = np.asarray(
                 [scale_factor[0] * im_scale_y, scale_factor[1] * im_scale_x],
                 dtype=np.float32)
         else:
-            sample['scale_factor'] = np.array(
+            sample['scale_factor'] = np.asarray(
                 [im_scale_y, im_scale_x], dtype=np.float32)
 
         # apply bbox
@@ -1397,8 +1397,8 @@ class RandomScaledCropOp(BaseOperator):
         random_dim = int(dim * random_scale)
         dim_max = max(h, w)
         scale = random_dim / dim_max
-        resize_w = int(round(w * scale))
-        resize_h = int(round(h * scale))
+        resize_w = w * scale
+        resize_h = h * scale
         offset_x = int(max(0, np.random.uniform(0., resize_w - dim)))
         offset_y = int(max(0, np.random.uniform(0., resize_h - dim)))
 
@@ -1408,9 +1408,9 @@ class RandomScaledCropOp(BaseOperator):
         canvas[:min(dim, resize_h), :min(dim, resize_w), :] = img[
             offset_y:offset_y + dim, offset_x:offset_x + dim, :]
         sample['image'] = canvas
-        sample['im_shape'] = np.array([resize_h, resize_w], dtype=np.int32)
+        sample['im_shape'] = np.asarray([resize_h, resize_w], dtype=np.float32)
         scale_factor = sample['sacle_factor']
-        sample['scale_factor'] = np.array(
+        sample['scale_factor'] = np.asarray(
             [scale_factor[0] * scale, scale_factor[1] * scale],
             dtype=np.float32)
 

ppdet/modeling/architecture/mask_rcnn.py

@@ -96,7 +96,8 @@ class MaskRCNN(BaseArch):
                 self.bbox_head_out, rois)
             # Refine bbox by the output from bbox_head at test stage
             self.bboxes = self.bbox_post_process(bbox_pred, bboxes,
-                                                 self.inputs['im_info'])
+                                                 self.inputs['im_shape'],
+                                                 self.inputs['scale_factor'])
         else:
             # Proposal RoI for Mask branch
             # bboxes update at training stage only
@@ -134,7 +135,8 @@ class MaskRCNN(BaseArch):
 
     def get_pred(self, ):
         mask = self.mask_post_process(self.bboxes, self.mask_head_out,
-                                      self.inputs['im_info'])
+                                      self.inputs['im_shape'],
+                                      self.inputs['scale_factor'])
         bbox, bbox_num = self.bboxes
         output = {
             'bbox': bbox.numpy(),

ppdet/modeling/bbox.py

@@ -93,6 +93,11 @@ class Proposal(object):
         self.proposal_target_generator = proposal_target_generator
 
     def generate_proposal(self, inputs, rpn_head_out, anchor_out):
+        # TODO: delete im_info 
+        try:
+            im_shape = inputs['im_info']
+        except:
+            im_shape = inputs['im_shape']
         rpn_rois_list = []
         rpn_prob_list = []
         rpn_rois_num_list = []
@@ -104,7 +109,7 @@ class Proposal(object):
                 bbox_deltas=rpn_delta,
                 anchors=anchor,
                 variances=var,
-                im_info=inputs['im_info'],
+                im_shape=im_shape,
                 mode=inputs['mode'])
             if len(rpn_head_out) == 1:
                 return rpn_rois, rpn_rois_num

ppdet/modeling/head/mask_head.py

@@ -138,7 +138,7 @@ class MaskHead(Layer):
         return mask_head_out
 
     def forward_test(self,
-                     im_info,
+                     scale_factor,
                      body_feats,
                      bboxes,
                      bbox_feat,
@@ -149,12 +149,14 @@ class MaskHead(Layer):
         if bbox.shape[0] == 0:
             mask_head_out = bbox
         else:
-            im_info_expand = []
+            scale_factor_list = []
             for idx, num in enumerate(bbox_num):
                 for n in range(num):
-                    im_info_expand.append(im_info[idx, -1])
-            im_info_expand = paddle.concat(im_info_expand)
-            scaled_bbox = paddle.multiply(bbox[:, 2:], im_info_expand, axis=0)
+                    scale_factor_list.append(scale_factor[idx, 0])
+            scale_factor_list = paddle.cast(
+                paddle.concat(scale_factor_list), 'float32')
+            scaled_bbox = paddle.multiply(
+                bbox[:, 2:], scale_factor_list, axis=0)
             scaled_bboxes = (scaled_bbox, bbox_num)
             mask_feat = self.mask_feat(body_feats, scaled_bboxes, bbox_feat,
                                        mask_index, spatial_scale, stage)
@@ -174,8 +176,8 @@ class MaskHead(Layer):
             mask_head_out = self.forward_train(body_feats, bboxes, bbox_feat,
                                                mask_index, spatial_scale, stage)
         else:
-            im_info = inputs['im_info']
-            mask_head_out = self.forward_test(im_info, body_feats, bboxes,
+            scale_factor = inputs['scale_factor']
+            mask_head_out = self.forward_test(scale_factor, body_feats, bboxes,
                                               bbox_feat, mask_index,
                                               spatial_scale, stage)
         return mask_head_out

ppdet/modeling/layers.py

@@ -16,8 +16,7 @@ import numpy as np
 from numbers import Integral
 
 import paddle
-import paddle.fluid as fluid
-from paddle.fluid.dygraph.base import to_variable
+from paddle import to_tensor
 from ppdet.core.workspace import register, serializable
 from ppdet.py_op.target import generate_rpn_anchor_target, generate_proposal_target, generate_mask_target
 from ppdet.py_op.post_process import bbox_post_process
@@ -86,20 +85,20 @@ class AnchorTargetGeneratorRPN(object):
             self.batch_size_per_im, self.positive_overlap,
             self.negative_overlap, self.fg_fraction, self.use_random)
 
-        loc_indexes = to_variable(loc_indexes)
-        score_indexes = to_variable(score_indexes)
-        tgt_labels = to_variable(tgt_labels)
-        tgt_bboxes = to_variable(tgt_bboxes)
-        bbox_inside_weights = to_variable(bbox_inside_weights)
+        loc_indexes = to_tensor(loc_indexes)
+        score_indexes = to_tensor(score_indexes)
+        tgt_labels = to_tensor(tgt_labels)
+        tgt_bboxes = to_tensor(tgt_bboxes)
+        bbox_inside_weights = to_tensor(bbox_inside_weights)
 
         loc_indexes.stop_gradient = True
         score_indexes.stop_gradient = True
         tgt_labels.stop_gradient = True
 
-        cls_logits = fluid.layers.reshape(x=cls_logits, shape=(-1, ))
-        bbox_pred = fluid.layers.reshape(x=bbox_pred, shape=(-1, 4))
-        pred_cls_logits = fluid.layers.gather(cls_logits, score_indexes)
-        pred_bbox_pred = fluid.layers.gather(bbox_pred, loc_indexes)
+        cls_logits = paddle.reshape(x=cls_logits, shape=(-1, ))
+        bbox_pred = paddle.reshape(x=bbox_pred, shape=(-1, 4))
+        pred_cls_logits = paddle.gather(cls_logits, score_indexes)
+        pred_bbox_pred = paddle.gather(bbox_pred, loc_indexes)
 
         return pred_cls_logits, pred_bbox_pred, tgt_labels, tgt_bboxes, bbox_inside_weights
 
@@ -131,14 +130,30 @@ class ProposalGenerator(object):
                  bbox_deltas,
                  anchors,
                  variances,
-                 im_info,
+                 im_shape,
                  mode='train'):
         pre_nms_top_n = self.train_pre_nms_top_n if mode == 'train' else self.infer_pre_nms_top_n
         post_nms_top_n = self.train_post_nms_top_n if mode == 'train' else self.infer_post_nms_top_n
+        # TODO delete im_info
+        if im_shape.shape[1] > 2:
+            import paddle.fluid as fluid
             rpn_rois, rpn_rois_prob, rpn_rois_num = fluid.layers.generate_proposals(
                 scores,
                 bbox_deltas,
-            im_info,
+                im_shape,
+                anchors,
+                variances,
+                pre_nms_top_n=pre_nms_top_n,
+                post_nms_top_n=post_nms_top_n,
+                nms_thresh=self.nms_thresh,
+                min_size=self.min_size,
+                eta=self.eta,
+                return_rois_num=True)
+        else:
+            rpn_rois, rpn_rois_prob, rpn_rois_num = ops.generate_proposals(
+                scores,
+                bbox_deltas,
+                im_shape,
                 anchors,
                 variances,
                 pre_nms_top_n=pre_nms_top_n,
@@ -198,7 +213,7 @@ class ProposalTargetGenerator(object):
             self.bg_thresh_hi[stage], self.bg_thresh_lo[stage],
             self.bbox_reg_weights[stage], self.num_classes, self.use_random,
             self.is_cls_agnostic, self.is_cascade_rcnn)
-        outs = [to_variable(v) for v in outs]
+        outs = [to_tensor(v) for v in outs]
         for v in outs:
             v.stop_gradient = True
         return outs
@@ -227,7 +242,7 @@ class MaskTargetGenerator(object):
                                     rois, rois_num, labels_int32,
                                     self.num_classes, self.mask_resolution)
 
-        outs = [to_variable(v) for v in outs]
+        outs = [to_tensor(v) for v in outs]
         for v in outs:
             v.stop_gradient = True
         return outs
@@ -260,7 +275,7 @@ class RCNNBox(object):
         scale_list = []
         origin_shape_list = []
         for idx in range(self.batch_size):
-            scale = scale_factor[idx, :]
+            scale = scale_factor[idx, :][0]
             rois_num_per_im = rois_num[idx]
             expand_scale = paddle.expand(scale, [rois_num_per_im, 1])
             scale_list.append(expand_scale)
@@ -327,7 +342,7 @@ class DecodeClipNms(object):
                                  im_info.numpy(), self.keep_top_k,
                                  self.score_threshold, self.nms_threshold,
                                  self.num_classes)
-        outs = [to_variable(v) for v in outs]
+        outs = [to_tensor(v) for v in outs]
         for v in outs:
             v.stop_gradient = True
         return outs
@@ -407,7 +422,6 @@ class YOLOBox(object):
     def __call__(self, yolo_head_out, anchors, im_shape, scale_factor=None):
         boxes_list = []
         scores_list = []
-        im_shape = paddle.cast(im_shape, 'float32')
         if scale_factor is not None:
             origin_shape = im_shape / scale_factor
         else:

ppdet/modeling/post_process.py

@@ -17,14 +17,7 @@ class BBoxPostProcess(object):
         self.nms = nms
 
     def __call__(self, head_out, rois, im_shape, scale_factor=None):
-        # TODO: compatible for im_info
-        # remove after unify the im_shape. scale_factor
-        if im_shape.shape[1] > 2:
-            origin_shape = im_shape[:, :2]
-            scale_factor = im_shape[:, 2:]
-        else:
-            origin_shape = im_shape
-        bboxes, score = self.decode(head_out, rois, origin_shape, scale_factor)
+        bboxes, score = self.decode(head_out, rois, im_shape, scale_factor)
         bbox_pred, bbox_num = self.nms(bboxes, score)
         return bbox_pred, bbox_num
 
@@ -38,12 +31,12 @@ class MaskPostProcess(object):
         self.mask_resolution = mask_resolution
         self.binary_thresh = binary_thresh
 
-    def __call__(self, bboxes, mask_head_out, im_info):
+    def __call__(self, bboxes, mask_head_out, im_shape, scale_factor=None):
         # TODO: modify related ops for deploying
         bboxes_np = (i.numpy() for i in bboxes)
         mask = mask_post_process(bboxes_np,
                                  mask_head_out.numpy(),
-                                 im_info.numpy(), self.mask_resolution,
-                                 self.binary_thresh)
+                                 im_shape.numpy(), scale_factor[:, 0].numpy(),
+                                 self.mask_resolution, self.binary_thresh)
         mask = {'mask': mask}
         return mask

ppdet/py_op/post_process.py

@@ -10,7 +10,8 @@ import cv2
 def bbox_post_process(bboxes,
                       bbox_prob,
                       bbox_deltas,
-                      im_info,
+                      im_shape,
+                      scale_factor,
                       keep_top_k=100,
                       score_thresh=0.05,
                       nms_thresh=0.5,
@@ -27,14 +28,14 @@ def bbox_post_process(bboxes,
         end_num += box_num
 
         boxes = bbox[st_num:end_num, :]  # bbox 
-        boxes = boxes / im_info[i][2]  # scale
+        boxes = boxes / scale_factor[i]  # scale
         bbox_delta = bbox_deltas[st_num:end_num, :, :]  # bbox delta 
         bbox_delta = np.reshape(bbox_delta, (box_num, -1))
         # step1: decode 
         boxes = delta2bbox(bbox_delta, boxes, bbox_reg_weights)
 
         # step2: clip 
-        boxes = clip_bbox(boxes, im_info[i][:2] / im_info[i][2])
+        boxes = clip_bbox(boxes, im_shape[i][:2] / scale_factor[i])
         # step3: nms 
         cls_boxes = [[] for _ in range(class_nums)]
         scores_n = bbox_prob[st_num:end_num, :]
@@ -72,7 +73,12 @@ def bbox_post_process(bboxes,
 
 
 @jit
-def mask_post_process(bboxes, masks, im_info, resolution=14, binary_thresh=0.5):
+def mask_post_process(bboxes,
+                      masks,
+                      im_shape,
+                      scale_factor,
+                      resolution=14,
+                      binary_thresh=0.5):
     if masks.shape[0] == 0:
         return masks
     bbox, bbox_nums = bboxes
@@ -93,8 +99,8 @@ def mask_post_process(bboxes, masks, im_info, resolution=14, binary_thresh=0.5):
         labels_n = labels[st_num:end_num]
         masks_n = masks[st_num:end_num]
 
-        im_h = int(round(im_info[i][0] / im_info[i][2]))
-        im_w = int(round(im_info[i][1] / im_info[i][2]))
+        im_h = int(round(im_shape[i][0] / scale_factor[i]))
+        im_w = int(round(im_shape[i][1] / scale_factor[i]))
         boxes_n = expand_bbox(boxes_n, scale)
         boxes_n = boxes_n.astype(np.int32)
         padded_mask = np.zeros((M + 2, M + 2), dtype=np.float32)