Skip to content
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

MAP metric, fix metric for CUDA execution #673

Merged
merged 8 commits into from
Dec 15, 2021
Merged

MAP metric, fix metric for CUDA execution #673

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
b7205db
move and initialize tensors on the correct device (fix cuda)
Dec 10, 2021
81fe48a
remove condition for moving tensors, its done in .to()
Dec 11, 2021
97e4327
Merge branch 'master' into bug/map-devices
Borda Dec 12, 2021
777b72e
changelog
SkafteNicki Dec 13, 2021
41abc33
Merge branch 'master' into bug/map-devices
Borda Dec 13, 2021
2972ef0
Merge branch 'master' into bug/map-devices
Borda Dec 13, 2021
dd5680a
fix gpu test
SkafteNicki Dec 15, 2021
b0a798e
docs
Borda Dec 15, 2021
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
2 changes: 2 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,6 +56,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
- Fixed `torch.sort` currently does not support bool dtype on CUDA ([#665](https://github.com/PyTorchLightning/metrics/pull/665))


- Fixed initialization of tensors to be on correct device for `MAP` metric ([#673](https://github.com/PyTorchLightning/metrics/pull/673))


## [0.6.1] - 2021-12-06

Expand Down
10 changes: 7 additions & 3 deletions torchmetrics/detection/map.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -423,10 +423,10 @@ def _evaluate_image(
nb_iou_thrs = len(self.iou_thresholds)
nb_gt = len(gt)
nb_det = len(det)
gt_matches = torch.zeros((nb_iou_thrs, nb_gt), dtype=torch.bool)
det_matches = torch.zeros((nb_iou_thrs, nb_det), dtype=torch.bool)
gt_matches = torch.zeros((nb_iou_thrs, nb_gt), dtype=torch.bool, device=self.device)
det_matches = torch.zeros((nb_iou_thrs, nb_det), dtype=torch.bool, device=self.device)
gt_ignore = ignore_area_sorted
det_ignore = torch.zeros((nb_iou_thrs, nb_det), dtype=torch.bool)
det_ignore = torch.zeros((nb_iou_thrs, nb_det), dtype=torch.bool, device=self.device)
if torch.numel(ious) > 0:
for idx_iou, t in enumerate(self.iou_thresholds):
for idx_det in range(nb_det):
Expand Down Expand Up @@ -565,6 +565,10 @@ def _calculate(self, class_ids: List) -> Tuple[Dict, MAPMetricResults, MARMetric
recall = -torch.ones((nb_iou_thrs, nb_classes, nb_bbox_areas, nb_max_det_thrs))
scores = -torch.ones((nb_iou_thrs, nb_rec_thrs, nb_classes, nb_bbox_areas, nb_max_det_thrs))

# move tensors if necessary
self.max_detection_thresholds = self.max_detection_thresholds.to(self.device)
self.rec_thresholds = self.rec_thresholds.to(self.device)

# retrieve E at each category, area range, and max number of detections
for idx_cls in range(nb_classes):
for idx_bbox_area in range(nb_bbox_areas):
Expand Down