[MaskRCNN] pack gt masks to bits

examples/FasterRCNN/data.py

@@ -179,13 +179,15 @@ class TrainingDataPreprocessor:
             # And produce one image-sized binary mask per box.
             masks = []
             width_height = np.asarray([width, height], dtype=np.float32)
+            gt_mask_width = int(np.ceil(im.shape[1] / 8.0) * 8)   # pad to 8 in order to pack mask into bits
             for polys in segmentation:
                 if not self.cfg.DATA.ABSOLUTE_COORD:
                     polys = [p * width_height for p in polys]
                 polys = [self.aug.augment_coords(p, params) for p in polys]
-                masks.append(segmentation_to_mask(polys, im.shape[0], im.shape[1]))
+                masks.append(segmentation_to_mask(polys, im.shape[0], gt_mask_width))
             masks = np.asarray(masks, dtype='uint8')    # values in {0, 1}
-            ret['gt_masks'] = masks
+            masks = np.packbits(masks, axis=-1)
+            ret['gt_masks_packed'] = masks
 
             # from viz import draw_annotation, draw_mask
             # viz = draw_annotation(im, boxes, klass)

examples/FasterRCNN/modeling/generalized_rcnn.py

@@ -20,7 +20,7 @@ from .model_cascade import CascadeRCNNHead
 from .model_fpn import fpn_model, generate_fpn_proposals, multilevel_roi_align, multilevel_rpn_losses
 from .model_frcnn import (
     BoxProposals, FastRCNNHead, fastrcnn_outputs, fastrcnn_predictions, sample_fast_rcnn_targets)
-from .model_mrcnn import maskrcnn_loss, maskrcnn_upXconv_head
+from .model_mrcnn import maskrcnn_loss, maskrcnn_upXconv_head, unpackbits_masks
 from .model_rpn import generate_rpn_proposals, rpn_head, rpn_losses
 
 
@@ -62,6 +62,9 @@ class GeneralizedRCNN(ModelDesc):
 
     def build_graph(self, *inputs):
         inputs = dict(zip(self.input_names, inputs))
+        if "gt_masks_packed" in inputs:
+            gt_masks = tf.cast(unpackbits_masks(inputs.pop("gt_masks_packed")), tf.uint8, name="gt_masks")
+            inputs["gt_masks"] = gt_masks
 
         image = self.preprocess(inputs['image'])     # 1CHW
 
@@ -91,8 +94,8 @@ class ResNetC4Model(GeneralizedRCNN):
             tf.TensorSpec((None,), tf.int64, 'gt_labels')]  # all > 0
         if cfg.MODE_MASK:
             ret.append(
-                tf.TensorSpec((None, None, None), tf.uint8, 'gt_masks')
-            )   # NR_GT x height x width
+                tf.TensorSpec((None, None, None), tf.uint8, 'gt_masks_packed')
+            )   # NR_GT x height x ceil(width/8), packed groundtruth masks
         return ret
 
     def backbone(self, image):
@@ -202,8 +205,8 @@ class ResNetFPNModel(GeneralizedRCNN):
             tf.TensorSpec((None,), tf.int64, 'gt_labels')])  # all > 0
         if cfg.MODE_MASK:
             ret.append(
-                tf.TensorSpec((None, None, None), tf.uint8, 'gt_masks')
-            )   # NR_GT x height x width
+                tf.TensorSpec((None, None, None), tf.uint8, 'gt_masks_packed')
+            )
         return ret
 
     def slice_feature_and_anchors(self, p23456, anchors):

examples/FasterRCNN/modeling/model_mrcnn.py

@@ -85,3 +85,22 @@ def maskrcnn_up4conv_head(*args, **kwargs):
 
 def maskrcnn_up4conv_gn_head(*args, **kwargs):
     return maskrcnn_upXconv_head(*args, num_convs=4, norm='GN', **kwargs)
+
+
+def unpackbits_masks(masks):
+    """
+    Args:
+        masks (Tensor): uint8 Tensor of shape N, H, W. The last dimension is packed bits.
+
+    Returns:
+        masks (Tensor): bool Tensor of shape N, H, 8*W.
+
+    This is a reverse operation of `np.packbits`
+    """
+    assert masks.dtype == tf.uint8, masks
+    bits = tf.constant((128, 64, 32, 16, 8, 4, 2, 1), dtype=tf.uint8)
+    unpacked = tf.bitwise.bitwise_and(tf.expand_dims(masks, -1), bits) > 0
+    unpacked = tf.reshape(
+        unpacked,
+        tf.concat([tf.shape(masks)[:-1], [-1]], axis=0))
+    return unpacked

tensorpack/utils/concurrency.py

@@ -238,7 +238,7 @@ def start_proc_mask_signal(proc):
                 if sys.version_info < (3, 4) or mp.get_start_method() == 'fork':
                     log_once(
 "Starting a process with 'fork' method is not safe and may consume unnecessary extra memory."
-" Use 'forkserver' method (available after Py3.4) instead if you run into any issues. "
+" Use 'forkserver/spawn' method (available after Py3.4) instead if you run into any issues. "
 "See https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods",
 'warn')  # noqa
             p.start()