roialign zero output

mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/roi_align_impl.cu

@@ -91,7 +91,7 @@ __device__ void bin_box(int thread_idx, const T *roi_boxes, int roi_cols, const
   }
 
   // Scale and shift ROI
-  T roi_offset = roi_end_mode == 1 ? static_cast<T>(0.5) : static_cast<T>(.0);
+  T roi_offset = roi_end_mode == 0 ? static_cast<T>(0.5) : static_cast<T>(.0);
   *roi_start_w = roi_box[0] * spatial_scale - roi_offset;
   *roi_start_h = roi_box[1] * spatial_scale - roi_offset;
   T roi_end_w = roi_box[2] * spatial_scale - roi_offset;
@@ -121,10 +121,9 @@ __global__ void ROIAlignKernel(size_t size, const T *input, const T *roi_boxes,
        thread_idx += blockDim.x * gridDim.x) {
     int n = thread_idx / pooled_width / pooled_height / channels;
     const T *roi_box = roi_boxes + n * roi_cols;
-    if (roi_box[0] < static_cast<T>(0.001) && roi_box[1] < static_cast<T>(0.001) &&
-        roi_box[2] < static_cast<T>(0.001) && roi_box[3] < static_cast<T>(0.001) &&
-        roi_box[0] > static_cast<T>(-0.001) && roi_box[1] > static_cast<T>(-0.001) &&
-        roi_box[2] > static_cast<T>(-0.001) && roi_box[3] > static_cast<T>(-0.001)) {
+    // Skip if roi box is a line
+    if (roi_box[1] < static_cast<T>(0.001) && roi_box[3] < static_cast<T>(0.001) &&
+        roi_box[1] > static_cast<T>(-0.001) && roi_box[3] > static_cast<T>(-0.001)) {
       continue;
     }
 
@@ -136,8 +135,6 @@ __global__ void ROIAlignKernel(size_t size, const T *input, const T *roi_boxes,
             pooled_height, pooled_width, &offset, &n, &c, &ph, &pw, &roi_bin_grid_h, &roi_bin_grid_w, &bin_size_h,
             &bin_size_w, &roi_start_h, &roi_start_w);
 
-    if (offset < 0 || offset >= size) continue;
-
     // (n, c, ph, pw) is the base param of pooled map
     const T count_points_in_grid_cell = roi_bin_grid_h * roi_bin_grid_w;
 
@@ -209,10 +206,8 @@ __global__ void ROIAlignGradKernel(size_t size, const T *dy, const T *roi_boxes,
        thread_idx += blockDim.x * gridDim.x) {
     int n = thread_idx / pooled_width / pooled_height / channels;
     const T *roi_box = roi_boxes + n * roi_cols;
-    if (roi_box[0] < static_cast<T>(0.001) && roi_box[1] < static_cast<T>(0.001) &&
-        roi_box[2] < static_cast<T>(0.001) && roi_box[3] < static_cast<T>(0.001) &&
-        roi_box[0] > static_cast<T>(-0.001) && roi_box[1] > static_cast<T>(-0.001) &&
-        roi_box[2] > static_cast<T>(-0.001) && roi_box[3] > static_cast<T>(-0.001)) {
+    if (roi_box[1] < static_cast<T>(0.001) && roi_box[3] < static_cast<T>(0.001) &&
+        roi_box[1] > static_cast<T>(-0.001) && roi_box[3] > static_cast<T>(-0.001)) {
       continue;
     }
 
@@ -224,8 +219,6 @@ __global__ void ROIAlignGradKernel(size_t size, const T *dy, const T *roi_boxes,
             pooled_height, pooled_width, &offset, &n, &c, &ph, &pw, &roi_bin_grid_h, &roi_bin_grid_w, &bin_size_h,
             &bin_size_w, &roi_start_h, &roi_start_w);
 
-    if (offset < 0 || offset >= size) continue;
-
     // (n, c, ph, pw) is the base param of pooled map
     const T count_points_in_grid_cell = roi_bin_grid_h * roi_bin_grid_w;
 

tests/st/ops/gpu/test_roi_align_grad_half_op.py

@@ -62,10 +62,17 @@ def test_roi_align_grad_half():
         sample_num)
     output = roi_align_grad(dy, rois)
     print(output)
-    expect = ([[[[0.0563, 0.0563, 0.0750, 0.0938, 0.1125, 0.0563],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0188, 0.0188, 0.0250, 0.0312, 0.0375, 0.0188]]]])
+    # the out if aligned is True
+    # expect = ([[[[0.0563, 0.0563, 0.0750, 0.0938, 0.1125, 0.0563],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0188, 0.0188, 0.0250, 0.0312, 0.0375, 0.0188]]]])
+    expect = ([[[[0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075]]]])
     np.testing.assert_almost_equal(output.asnumpy(), expect, decimal=4)

tests/st/ops/gpu/test_roi_align_grad_op.py

@@ -62,10 +62,17 @@ def test_roi_align_grad():
         sample_num)
     output = roi_align_grad(dy, rois)
     print(output)
-    expect = ([[[[0.0563, 0.0563, 0.0750, 0.0938, 0.1125, 0.0563],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
-                 [0.0188, 0.0188, 0.0250, 0.0312, 0.0375, 0.0188]]]])
+    # the out if aligned is True
+    # expect = ([[[[0.0563, 0.0563, 0.0750, 0.0938, 0.1125, 0.0563],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0375, 0.0375, 0.0500, 0.0625, 0.0750, 0.0375],
+    #              [0.0188, 0.0188, 0.0250, 0.0312, 0.0375, 0.0188]]]])
+    expect = ([[[[0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075],
+                 [0.025, 0.025, 0.05, 0.05, 0.075, 0.075]]]])
     np.testing.assert_almost_equal(output.asnumpy(), expect, decimal=4)

tests/st/ops/gpu/test_roi_align_half_op.py

@@ -39,7 +39,7 @@ def test_roi_align_half():
 
     # test case 1
     pooled_height, pooled_width, spatial_scale, sample_num = 4, 4, 0.2, 3
-    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num)
+    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num, 0)
     output = roi_align(x, rois)
     print(output)
     expect = [[[[1.2333, 2.1000, 3.3000, 4.5000],

tests/st/ops/gpu/test_roi_align_op.py

@@ -39,7 +39,7 @@ def test_roi_align():
 
     # test case 1
     pooled_height, pooled_width, spatial_scale, sample_num = 3, 3, 0.25, 2
-    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num)
+    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num, 0)
     output = roi_align(x, rois)
     print(output)
     expect = [[[[2.75, 4.5, 6.5],
@@ -49,7 +49,7 @@ def test_roi_align():
 
     # test case 2
     pooled_height, pooled_width, spatial_scale, sample_num = 4, 4, 0.2, 3
-    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num)
+    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num, 0)
     output = roi_align(x, rois)
     print(output)
     expect = [[[[1.2333, 2.1000, 3.3000, 4.5000],
@@ -63,7 +63,7 @@ def test_roi_align():
     rois = Tensor(np.array([[0, -2.0, -2.0, 22.0, 22.0],
                             [0, 1.0, 0.0, 19.0, 18.0]],
                            np.float32))
-    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num)
+    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num, 0)
     output = roi_align(x, rois)
     print(output)
     expect = [[[[3.3333, 5.5000, 7.6667],
@@ -77,7 +77,7 @@ def test_roi_align():
     # test case 4
     pooled_height, pooled_width, spatial_scale, sample_num = 2, 2, 1.0, -1
     rois = Tensor(np.array([[0, -2.0, -2.0, 22.0, 22.0]], np.float32))
-    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num)
+    roi_align = P.ROIAlign(pooled_height, pooled_width, spatial_scale, sample_num, 0)
     output = roi_align(x, rois)
     print(output)
     expect = [[[[8.2222, 0.],