[bugfix] set default max_dets to same values as pycocotools #9105
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Motivation
The COCO dataset recall evaluation in mmdetection differs from the original pycocotools. This is because the mmdetection's default
maxDets
([100, 500, 1000]
) are not equal to the others ([1, 10, 100]
).This change fixes issue #6884.
Modification
proposal_nums
tomax_dets
proposal_nums
seems to be used in the first stage of two-stage object detection (many proposals will appear).max_dets
is compatible to coco evaluation.cocoEval.params.maxDets
from[100, 500, 1000]
to[1, 10, 100]
[100, 500, 1000]
cocoEval.params.maxDets
only ifmetric == "proposal"
. I didn't because you'll change this variable in other situations (such as detection of a large number of objects).BC-breaking (Optional)
maxDets=[100, 500, 1000]
or simply see recall@100 (because recall@[100, 500, 1000] will get the same value in most experiments).Use cases (Optional)
COCO evaluation
Checklist