Skip to content

[PT FE] Torchvision NMS can accept negative scores #19826

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Sep 14, 2023
Merged

[PT FE] Torchvision NMS can accept negative scores #19826

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions src/frontends/pytorch/src/op/nms.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,8 +34,10 @@ OutputVector translate_nms(const NodeContext& context) {
context.mark_node(v0::Constant::create(element::i32, Shape{1}, {std::numeric_limits<int32_t>::max()}));
auto iou_threshold = context.get_input(2);

auto nms_out =
context.mark_node(std::make_shared<v9::NonMaxSuppression>(boxes, scores, max_output_per_class, iou_threshold));
auto score_threshold =
context.mark_node(v0::Constant::create(element::f32, Shape{}, {std::numeric_limits<float>::lowest()}));
auto nms_out = context.mark_node(
std::make_shared<v9::NonMaxSuppression>(boxes, scores, max_output_per_class, iou_threshold, score_threshold));
auto select = context.mark_node(std::make_shared<v8::Gather>(nms_out, const_2, const_1));

return {context.mark_node(std::make_shared<v0::Squeeze>(select, const_1))};
Expand Down
3 changes: 2 additions & 1 deletion tests/layer_tests/pytorch_tests/test_nms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,8 @@ def _prepare_input(self):
# PyTorch requires that boxes are in (x1, y1, x2, y2) format, where 0<=x1<x2 and 0<=y1<y2
boxes = np.array([[np.random.uniform(1, 3), np.random.uniform(2, 6),
np.random.uniform(4, 6), np.random.uniform(7, 9)] for _ in range(self.boxes_num)]).astype(np.float32)
scores = np.abs(np.random.randn(self.boxes_num).astype(np.float32))
# scores can be negative
scores = np.random.randn(self.boxes_num).astype(np.float32)
return (boxes, scores)

def create_model(self):
Expand Down