experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add an option to clip/not clip box outputs in CombinedNonMaxSuppression.

Browse Source 
PiperOrigin-RevId: 243104621
This commit is contained in:
Pengchong Jin 2019-04-11 11:13:09 -07:00 committed by TensorFlower Gardener
parent 9d724a8e60
commit 35e39cd3f1
12 changed files with 78 additions and 31 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
tensorflow/core/api_def/base_api/api_def_CombinedNonMaxSuppression.pbtxt
View File

@ -50,6 +50,14 @@ are padded/clipped to `max_total_size`. If true, the
output nmsed boxes, scores and classes are padded to be of length
`max_size_per_class`*`num_classes`, unless it exceeds `max_total_size` in
which case it is clipped to `max_total_size`. Defaults to false.
END
}
attr {
name: "clip_boxes"
description: <<END
If true, assume the box coordinates are between [0, 1] and clip the output boxes
if they fall beyond [0, 1]. If false, do not do clipping and output the box
coordinates as it is.
END
}
out_arg {

4
tensorflow/core/api_def/python_api/api_def_CombinedNonMaxSuppression.pbtxt Normal file
View File

@ -0,0 +1,4 @@
op {
graph_op_name: "CombinedNonMaxSuppression"
visibility: HIDDEN
}

30
tensorflow/core/kernels/non_max_suppression_op.cc
View File

@ -227,14 +227,11 @@ void BatchedNonMaxSuppressionOp(
OpKernelContext* context, const Tensor& inp_boxes, const Tensor& inp_scores,
int num_boxes, const int max_size_per_class, const int total_size_per_batch,
const float score_threshold, const float iou_threshold,
bool pad_per_class = false) {
bool pad_per_class = false, bool clip_boxes = true) {
int q = inp_boxes.dim_size(2);
int num_classes = inp_scores.dim_size(2);
const int num_batches = inp_boxes.dim_size(0);
// Default clip window of [0, 0, 1, 1] if none specified
std::vector<float> clip_window{0, 0, 1, 1};
// [num_batches, per_batch_size * 4]
std::vector<std::vector<float>> nmsed_boxes(num_batches);
// [num_batches, per_batch_size]
@ -375,18 +372,23 @@ void BatchedNonMaxSuppressionOp(
while (curr_total_size > 0 && result_idx < result_candidate_vec.size()) {
ResultCandidate next_candidate = result_candidate_vec[result_idx++];
// Add to final output vectors
if (clip_boxes) {
const float box_min = 0.0;
const float box_max = 1.0;
nmsed_boxes[batch].push_back(
std::max(std::min(next_candidate.box_coord[0], clip_window[2]),
clip_window[0]));
std::max(std::min(next_candidate.box_coord[0], box_max), box_min));
nmsed_boxes[batch].push_back(
std::max(std::min(next_candidate.box_coord[1], clip_window[3]),
clip_window[1]));
std::max(std::min(next_candidate.box_coord[1], box_max), box_min));
nmsed_boxes[batch].push_back(
std::max(std::min(next_candidate.box_coord[2], clip_window[2]),
clip_window[0]));
std::max(std::min(next_candidate.box_coord[2], box_max), box_min));
nmsed_boxes[batch].push_back(
std::max(std::min(next_candidate.box_coord[3], clip_window[3]),
clip_window[1]));
std::max(std::min(next_candidate.box_coord[3], box_max), box_min));
} else {
nmsed_boxes[batch].push_back(next_candidate.box_coord[0]);
nmsed_boxes[batch].push_back(next_candidate.box_coord[1]);
nmsed_boxes[batch].push_back(next_candidate.box_coord[2]);
nmsed_boxes[batch].push_back(next_candidate.box_coord[3]);
}
nmsed_scores[batch].push_back(next_candidate.score);
nmsed_classes[batch].push_back(next_candidate.class_idx);
curr_total_size--;
@ -679,6 +681,7 @@ class CombinedNonMaxSuppressionOp : public OpKernel {
explicit CombinedNonMaxSuppressionOp(OpKernelConstruction* context)
: OpKernel(context) {
OP_REQUIRES_OK(context, context->GetAttr("pad_per_class", &pad_per_class_));
OP_REQUIRES_OK(context, context->GetAttr("clip_boxes", &clip_boxes_));
}
void Compute(OpKernelContext* context) override {
@ -734,11 +737,12 @@ class CombinedNonMaxSuppressionOp : public OpKernel {
BatchedNonMaxSuppressionOp(context, boxes, scores, num_boxes,
max_size_per_class, max_total_size_per_batch,
score_threshold_val, iou_threshold_val,
pad_per_class_);
pad_per_class_, clip_boxes_);
}
private:
bool pad_per_class_;
bool clip_boxes_;
};
REGISTER_KERNEL_BUILDER(Name("NonMaxSuppression").Device(DEVICE_CPU),

36
tensorflow/core/kernels/non_max_suppression_op_test.cc
View File

@ -863,7 +863,7 @@ TEST_F(NonMaxSuppressionWithOverlapsOpTest, TestEmptyInput) {
class CombinedNonMaxSuppressionOpTest : public OpsTestBase {
protected:
void MakeOp(bool pad_per_class = false) {
void MakeOp(bool pad_per_class = false, bool clip_boxes = true) {
TF_EXPECT_OK(NodeDefBuilder("combined_non_max_suppression_op",
"CombinedNonMaxSuppression")
.Input(FakeInput(DT_FLOAT))
@ -873,6 +873,7 @@ class CombinedNonMaxSuppressionOpTest : public OpsTestBase {
.Input(FakeInput(DT_FLOAT))
.Input(FakeInput(DT_FLOAT))
.Attr("pad_per_class", pad_per_class)
.Attr("clip_boxes", clip_boxes)
.Finalize(node_def()));
TF_EXPECT_OK(InitOp());
}
@ -942,6 +943,39 @@ TEST_F(CombinedNonMaxSuppressionOpTest, TestSelectFromThreeClusters) {
test::ExpectTensorEqual<int>(expected_valid_d, *GetOutput(3));
}
TEST_F(CombinedNonMaxSuppressionOpTest,
TestSelectFromThreeClustersNoBoxClipping) {
MakeOp(false, false);
AddInputFromArray<float>(TensorShape({1, 6, 1, 4}),
{0, 0, 10, 10, 0, 1, 10, 11, 0, 1, 10, 9,
0, 11, 10, 20, 0, 12, 10, 21, 0, 30, 100, 40});
AddInputFromArray<float>(TensorShape({1, 6, 1}),
{.9f, .75f, .6f, .95f, .5f, .3f});
AddInputFromArray<int>(TensorShape({}), {3});
AddInputFromArray<int>(TensorShape({}), {3});
AddInputFromArray<float>(TensorShape({}), {.5f});
AddInputFromArray<float>(TensorShape({}), {0.0f});
TF_ASSERT_OK(RunOpKernel());
// boxes
Tensor expected_boxes(allocator(), DT_FLOAT, TensorShape({1, 3, 4}));
test::FillValues<float>(&expected_boxes,
{0, 11, 10, 20, 0, 0, 10, 10, 0, 30, 100, 40});
test::ExpectTensorEqual<float>(expected_boxes, *GetOutput(0));
// scores
Tensor expected_scores(allocator(), DT_FLOAT, TensorShape({1, 3}));
test::FillValues<float>(&expected_scores, {0.95, 0.9, 0.3});
test::ExpectTensorEqual<float>(expected_scores, *GetOutput(1));
// classes
Tensor expected_classes(allocator(), DT_FLOAT, TensorShape({1, 3}));
test::FillValues<float>(&expected_classes, {0, 0, 0});
test::ExpectTensorEqual<float>(expected_classes, *GetOutput(2));
// valid
Tensor expected_valid_d(allocator(), DT_INT32, TensorShape({1}));
test::FillValues<int>(&expected_valid_d, {3});
test::ExpectTensorEqual<int>(expected_valid_d, *GetOutput(3));
}
TEST_F(CombinedNonMaxSuppressionOpTest,
TestSelectFromThreeClustersWithScoreThreshold) {
MakeOp();

1
tensorflow/core/ops/image_ops.cc
View File

@ -922,6 +922,7 @@ REGISTER_OP("CombinedNonMaxSuppression")
.Output("nmsed_classes: float")
.Output("valid_detections: int32")
.Attr("pad_per_class: bool = false")
.Attr("clip_boxes: bool = true")
.SetShapeFn(CombinedNMSShapeFn);
} // namespace tensorflow

6
tensorflow/python/ops/image_ops_impl.py
View File

@ -3496,6 +3496,7 @@ def combined_non_max_suppression(boxes,
iou_threshold=0.5,
score_threshold=float('-inf'),
pad_per_class=False,
clip_boxes=True,
name=None):
"""Greedily selects a subset of bounding boxes in descending order of score.
@ -3532,6 +3533,9 @@ def combined_non_max_suppression(boxes,
scores and classes are padded to be of length
`max_size_per_class`*`num_classes`, unless it exceeds `max_total_size` in
which case it is clipped to `max_total_size`. Defaults to false.
clip_boxes: If true, the coordinates of output nmsed boxes will be clipped
to [0, 1]. If false, output the box coordinates as it is. Defaults to
true.
name: A name for the operation (optional).
Returns:
@ -3553,7 +3557,7 @@ def combined_non_max_suppression(boxes,
score_threshold, dtype=dtypes.float32, name='score_threshold')
return gen_image_ops.combined_non_max_suppression(
boxes, scores, max_output_size_per_class, max_total_size, iou_threshold,
score_threshold, pad_per_class)
score_threshold, pad_per_class, clip_boxes)
@tf_export('image.draw_bounding_boxes', v1=[])

2
tensorflow/tools/api/golden/v1/tensorflow.image.pbtxt
View File

@ -34,7 +34,7 @@ tf_module {
}
member_method {
name: "combined_non_max_suppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'0.5\', \'-inf\', \'False\', \'None\'], "
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'clip_boxes\', \'name\'], varargs=None, keywords=None, defaults=[\'0.5\', \'-inf\', \'False\', \'True\', \'None\'], "
}
member_method {
name: "convert_image_dtype"

4
tensorflow/tools/api/golden/v1/tensorflow.pbtxt
View File

@ -964,10 +964,6 @@ tf_module {
name: "colocate_with"
argspec: "args=[\'op\', \'ignore_existing\'], varargs=None, keywords=None, defaults=[\'False\'], "
}
member_method {
name: "combined_non_max_suppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'None\'], "
}
member_method {
name: "complex"
argspec: "args=[\'real\', \'imag\', \'name\'], varargs=None, keywords=None, defaults=[\'None\'], "

2
tensorflow/tools/api/golden/v1/tensorflow.raw_ops.pbtxt
View File

@ -618,7 +618,7 @@ tf_module {
}
member_method {
name: "CombinedNonMaxSuppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'None\'], "
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'clip_boxes\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'True\', \'None\'], "
}
member_method {
name: "CompareAndBitpack"

2
tensorflow/tools/api/golden/v2/tensorflow.image.pbtxt
View File

@ -34,7 +34,7 @@ tf_module {
}
member_method {
name: "combined_non_max_suppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'0.5\', \'-inf\', \'False\', \'None\'], "
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'clip_boxes\', \'name\'], varargs=None, keywords=None, defaults=[\'0.5\', \'-inf\', \'False\', \'True\', \'None\'], "
}
member_method {
name: "convert_image_dtype"

4
tensorflow/tools/api/golden/v2/tensorflow.pbtxt
View File

@ -508,10 +508,6 @@ tf_module {
name: "clip_by_value"
argspec: "args=[\'t\', \'clip_value_min\', \'clip_value_max\', \'name\'], varargs=None, keywords=None, defaults=[\'None\'], "
}
member_method {
name: "combined_non_max_suppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'None\'], "
}
member_method {
name: "complex"
argspec: "args=[\'real\', \'imag\', \'name\'], varargs=None, keywords=None, defaults=[\'None\'], "

2
tensorflow/tools/api/golden/v2/tensorflow.raw_ops.pbtxt
View File

@ -618,7 +618,7 @@ tf_module {
}
member_method {
name: "CombinedNonMaxSuppression"
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'None\'], "
argspec: "args=[\'boxes\', \'scores\', \'max_output_size_per_class\', \'max_total_size\', \'iou_threshold\', \'score_threshold\', \'pad_per_class\', \'clip_boxes\', \'name\'], varargs=None, keywords=None, defaults=[\'False\', \'True\', \'None\'], "
}
member_method {
name: "CompareAndBitpack"