Methods › Computer Vision › Proposal Filtering › Matrix NMS
Matrix Non-Maximum Suppression
Matrix NMS
Introduced by Xinlong Wang et al. in SOLOv2: Dynamic and Fast Instance Segmentation
archive 2025-07-28 Description, source and code snippet are the archive's method entry.
Matrix NMS, or Matrix Non-Maximum Suppression, performs non-maximum suppression with parallel matrix operations in one shot. It is motivated by Soft-NMS. Soft-NMS decays the other detection scores as a monotonic decreasing function f(iou) of their overlaps. By decaying the scores according to IoUs recursively, higher IoU detections will be eliminated with a minimum score threshold. However, such process is sequential like traditional Greedy NMS and can not be implemented in parallel.
Matrix NMS views this process from another perspective by considering how a predicted mask mⱼ being suppressed. For mⱼ, its decay factor is affected by: (a) The penalty of each prediction mᵢ on mⱼ (sᵢ>sⱼ), where sᵢ and sⱼ are the confidence scores; and (b) the probability of mᵢ being suppressed. For (a), the penalty of each prediction mᵢ on mⱼ could be easily computed by f(. iou ._(i, j)). For (b), the probability of mᵢ being suppressed is not so elegant to be computed. However, the probability usually has positive correlation with the IoUs. So here we directly approximate the probability by the most overlapped prediction on mᵢ as
f( iou. _(, i))=min_(∀sₖ>sᵢ) f( iou _(k, i))
To this end, the final decay factor becomes
decayⱼ=min_(∀sᵢ>sⱼ) (f( iou _(i, j)))/(f( iou _(·, i)))
and the updated score is computed by sⱼ=sⱼ · decay ⱼ . The authors consider the two most simple decremented functions, denoted as linear f(. iou ._(i, j))=1- iou _(i, j), and Gaussian f(. iou ._(i, j))=exp(-(i o u_(i, j)²)/σ).
In Syntology repository known, no verified implementation (declared_code; repositories: 5 listed of 18 in the graph, strongest link: official). This answer is about arXiv:2003.10152, the paper the archive credits with this method.
Code Syntology holds under other names, as the lookup lists them (code_nearby); not a complete list, and none of them is this method.
- Matrix Encoding Network (MatNet) · shares_name_tokens (shares name tokens with this name; a different method) · level V3 · retrieve with
get_reference_implementation(name='Matrix Encoding Network (MatNet)') - Truncated Matrix Power Iteration · shares_name_tokens (shares name tokens with this name; a different method) · level V3 · retrieve with
get_reference_implementation(name='Truncated Matrix Power Iteration') - matrix mechanism for the counting matrix · shares_name_tokens (shares name tokens with this name; a different method) · level V3 · retrieve with
get_reference_implementation(name='matrix mechanism for the counting matrix') - Adjacency Matrix-Aware (AMA) data formatting method · shares_name_tokens (shares name tokens with this name; a different method) · level V3 · retrieve with
get_reference_implementation(name='Adjacency Matrix-Aware (AMA) data formatting method')
Origin paper of this answer, arXiv:2003.10152 (its page on this site · its page on Syntology): Syntology's run record for its harvested code, 23 ran (of which 4 constructed an object rather than computing a result; 13 with no instrument failure: 1 honoured, 0 violated, 12 with no contract checked; 10 where Syntology's instrument failed) · 15 unverified (of 38 harvested samples) · 24 pointer-only (licence). The instrument figure counts failures of Syntology's instrument, not of the code. Ran means executed on a synthesized input, not that the code is correct.
Syntology's have check for this name (free, it answers coverage only), read 2026-10-05; ask it again live. A level is the served verification level of that one answer; nothing here is summed across methods.
What each level means, quoted from Syntology's server card: V0 “imports”; V1 “executes against a synthesized fixture and returns a value of the recorded type”; V2 “two implementations GENERATED from the paper's signature and spec agree numerically (1e-6) on fixture inputs” “neither of the two ever read the authors' own code, so V2 measures how unambiguously the paper specifies the method, not fidelity to the authors' implementation (measured 2026-09-11, commit 75eaa77: cross_checked_by is a second LLM on 2,628 of 2,831 served samples)”; V3 “paper-derived property tests pass on both of those same two generations.” The card adds: “No rung on this ladder compares the served code against the authors' implementation, so treat every level as evidence about runnability and internal agreement, not about matching the paper's results.” A level quoted in two parts is one definition the card writes as two clauses joined by a double hyphen, each clause quoted on its own (how these levels are quoted).
Papers archive 2025-07-28
archive papers tagged: 5 · with a code link: 4 · where Syntology ran a sample: 2 (2 with a run with no instrument failure, 0 where every run was a failure of Syntology's instrument) Syntology
Show: all tagged papersonly where code ran (2 of 5 tagged: 2 with a run with no instrument failure, 0 where every run was a failure of Syntology's instrument)
Page 1 of 1: papers 1 to 5 of 5, newest first by the archive's date (ties by slug), in archive order.
Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code, as “N ran (of which C constructed an object rather than computing a result; K with no instrument failure: H honoured, V violated, P with no contract checked; I where Syntology's instrument failed) · U unverified”; the instrument figure counts failures of Syntology's instrument, not of the code. It is per sample and not a correctness claim. When the archive marks a repository official for the paper, the line starts with that repository's state (the archive's flag, not a verdict on who wrote the code; “community repositories only” when every sample that ran came from a community repository, “official: no sample here; runs from other or unrecorded repositories” when some came from a repository the paper names or has in its text, or from none recorded); hover it for the repositories the samples that ran came from.
-
PP-YOLOE: An evolved version of YOLO 30 Mar 2022 · 8 repositories · arXiv:2203.16250Syntology official (archive's flag): 2 ran · 22 ran (of which 0 constructed an object rather than computing a result; 19 with no instrument failure: 2 honoured, 0 violated, 17 with no contract checked; 3 where Syntology's instrument failed) · 5 unverified (of 27 harvested samples) · 1 pointer-only (licence)
-
In Defense of Kalman Filtering for Polyp Tracking from Colonoscopy Videos 27 Jan 2022 · 0 repositories · arXiv:2201.11450
-
PP-YOLOv2: A Practical Object Detector 21 Apr 2021 · 1 repository · arXiv:2104.10419
-
PP-YOLO: An Effective and Efficient Implementation of Object Detector 23 Jul 2020 · 5 repositories · arXiv:2007.12099
-
SOLOv2: Dynamic and Fast Instance Segmentation 23 Mar 2020 · 18 repositories · arXiv:2003.10152Syntology community repositories only · 23 ran (of which 4 constructed an object rather than computing a result; 13 with no instrument failure: 1 honoured, 0 violated, 12 with no contract checked; 10 where Syntology's instrument failed) · 15 unverified (of 38 harvested samples) · 24 pointer-only (licence)
Tasks archive 2025-07-28
15 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.
Usage over time archive 2025-07-28
Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).
Categories archive 2025-07-28
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections