Make mrcnn_mask_target arg mask_size a 2d tuple

Signed-off-by: Serge Panev <[email protected]>

Author: Kh4L

src/operator/contrib/mrcnn_mask_target-inl.h

@@ -45,16 +45,17 @@ namespace mrcnn_index {
 struct MRCNNMaskTargetParam : public dmlc::Parameter<MRCNNMaskTargetParam> {
   int num_rois;
   int num_classes;
-  int mask_size;
   int sample_ratio;
+  mxnet::TShape mask_size;
 
   DMLC_DECLARE_PARAMETER(MRCNNMaskTargetParam) {
     DMLC_DECLARE_FIELD(num_rois)
     .describe("Number of sampled RoIs.");
     DMLC_DECLARE_FIELD(num_classes)
     .describe("Number of classes.");
     DMLC_DECLARE_FIELD(mask_size)
-    .describe("Size of the pooled masks.");
+    .set_expect_ndim(2).enforce_nonzero()
+    .describe("Size of the pooled masks height and width: (h, w).");
     DMLC_DECLARE_FIELD(sample_ratio).set_default(2)
     .describe("Sampling ratio of ROI align. Set to -1 to use adaptative size.");
   }
@@ -91,7 +92,8 @@ inline bool MRCNNMaskTargetShape(const NodeAttrs& attrs,
   CHECK_EQ(tshape[0], batch_size) << " batch size should be the same for all the inputs.";
 
   // out: 2 * (B, N, C, MS, MS)
-  auto oshape = Shape5(batch_size, num_rois, param.num_classes, param.mask_size, param.mask_size);
+  auto oshape = Shape5(batch_size, num_rois, param.num_classes,
+                       param.mask_size[0], param.mask_size[1]);
   out_shape->clear();
   out_shape->push_back(oshape);
   out_shape->push_back(oshape);

src/operator/contrib/mrcnn_mask_target.cu

@@ -196,15 +196,16 @@ __global__ void MRCNNMaskTargetKernel(const DType *rois,
                                       int num_gtmasks,
                                       int gt_height,
                                       int gt_width,
-                                      int mask_size,
+                                      int mask_size_h,
+                                      int mask_size_w,
                                       int sample_ratio) {
   // computing sampled_masks
   RoIAlignForward(gt_masks, rois, matches, total_out_el,
-                  num_classes, gt_height, gt_width, mask_size, mask_size,
+                  num_classes, gt_height, gt_width, mask_size_h, mask_size_w,
                   sample_ratio, num_rois, num_gtmasks, sampled_masks);
   // computing mask_cls
   int num_masks = batch_size * num_rois * num_classes;
-  int mask_vol = mask_size * mask_size;
+  int mask_vol = mask_size_h * mask_size_w;
   for (int mask_idx = blockIdx.x; mask_idx < num_masks; mask_idx += gridDim.x) {
     int cls_idx = mask_idx % num_classes;
     int roi_idx = (mask_idx / num_classes) % num_rois;
@@ -252,7 +253,8 @@ void MRCNNMaskTargetRun<gpu>(const MRCNNMaskTargetParam& param, const std::vecto
     (rois.dptr_, gt_masks.dptr_, matches.dptr_, cls_targets.dptr_,
     out_masks.dptr_, out_mask_cls.dptr_,
     num_el, batch_size, param.num_classes, param.num_rois,
-    num_gtmasks, gt_height, gt_width, param.mask_size, param.sample_ratio);
+    num_gtmasks, gt_height, gt_width,
+    param.mask_size[0], param.mask_size[1], param.sample_ratio);
     MSHADOW_CUDA_POST_KERNEL_CHECK(MRCNNMaskTargetKernel);
   });
 }

tests/python/unittest/test_contrib_operator.py

@@ -359,7 +359,7 @@ def test_op_mrcnn_mask_target():
 
     num_rois = 2
     num_classes = 4
-    mask_size = 3
+    mask_size = (3, 3)
     ctx = mx.gpu(0)
     # (B, N, 4)
     rois = mx.nd.array([[[2.3, 4.3, 2.2, 3.3],