v1.8.0

What's Changed

• Update pre-commit config to replace mdformat-black with mdformat-ruff by @Borda in #79
• fix(coco): warn on annotation id 0 ambiguous match sentinel by @Borda in #81
• fix(api): correct rleIou mismatch indexing by @Borda in #83
• Add GitHub Actions workflow for linting with pre-commit by @Borda in #85
• ci(tests): run optional dependency integrations by @Borda in #86
• fix(api): retain orphan annotations in indexes by @Borda in #87
• fix(mask): guard decoded RLE slices by @Borda in #88
• fix(utils): support gloo all gather by @Borda in #89
• fix(utils): preserve matrices during normalization by @Borda in #90
• test(tests): stabilize comparator and distributed checks by @Borda in #91
• ci(packaging): require Python 3.10 by @Borda in #92
• fix(core): address review findings in Python evaluation API by @Borda in #93
• fix(api): harden native RLE and evaluator inputs by @Borda in #94
• fix(utils): resolve evaluation regressions by @Borda in #95
• test(tests): harden medium review coverage by @Borda in #96
• chore(ci): resolve docs / package / CI review findings by @Borda in #97
• perf(api): release GIL during IoU compute [boost 2.59x] by @Borda in #104
• perf(api): avoid accumulation copies [boost 1.11x] by @Borda in #106
• perf(api): partition ignore ordering linearly [boost 3.77x] by @Borda in #107
• perf(api): parallelize image evaluation [boost 2.17x] by @Borda in #108
• perf(api): parallelize RLE IoU evaluation [boost 2.63x] by @Borda in #105
• perf(api): vectorize computeOks [boost 6.05x] by @Borda in #111
• perf(api): parallelize batch RLE ops [boost 2.80x] by @Borda in #113
• perf(api): hoist Python list length [boost 1.09x] by @Borda in #114
• perf(api): optimize native and annotation dispatch by @Borda in #112
• fix(core): harden low-level COCO API edges by @Borda in #116
• fix(cpp): harden native evaluator boundaries by @Borda in #117
• fix(utils): harden extra evaluation helpers by @Borda in #118
• add(tests): harden low-level regression coverage by @Borda in #119
• docs/ci: align low-level docs with verified behavior by @Borda in #120
• perf(coco): reuse sorted detections across maxDets by @Borda in #121
• perf(coco): parallelize category-area accumulation by @Borda in #122
• perf(coco): skip empty IoU pair calculations by @Borda in #123
• update version by @MiXaiLL76 in #124

New Contributors

• @dependabot[bot] made their first contribution in #99

Full Changelog: 1.7.2...1.8.0