Discofork
Repository brief

facebookresearch/detectron2

Read the upstream summary on the left, browse the cached forks below it, and load each fork comparison into the right-hand panel.

Cached analysis
cached 2026-03-31T09:51:46.573Z
1mo ago

facebookresearch/detectron2

Detectron2 is a widely used Facebook AI Research library for object detection, segmentation, and related visual recognition tasks. It is active, not archived, and has a large community footprint with 34,269 stars and 7,920 forks. The repo includes documentation, demos, model zoo material, configs, tests, tools, and project extensions, which makes it a practical base for forks that want to build on an established computer vision research stack.

GitHub
Loading tags...
Stars34,269
Forks7,920
Default branchmain
Last pushed2026-03-21T12:01:27Z
Recommended shortcuts

Jump straight into Discofork's strongest cached fork picks, or open a compare view in one click.

Forks

Choose a fork to inspect

10 of 10 fork briefs
Selected

Choose this fork if ResNeSt support is the priority. Avoid it if you want the newest Detectron2 behavior, broader backbone options, or low-maintenance upstream compatibility.

Choose this fork if bottom-up attention and its surrounding experiment stack matter more than staying current with upstream Detectron2. Avoid it if you want an actively maintained, general-purpose Detectron2 base.

Choose this fork if you need a document-layout-oriented Detectron2 variant and can tolerate major divergence and low maintenance. Choose upstream if you want current Detectron2 features, cleaner deployment/export paths, and easier long-term upkeep.

Prefer this fork if Windows support is the main requirement and you can tolerate being behind upstream. Prefer upstream if you need current model support, active maintenance, or the broadest Detectron2 feature set.

Choose this fork if Windows support is the main requirement and you can accept an old, heavily diverged codebase. Choose upstream if you want current Detectron2 features, active maintenance, and less upgrade risk.

Choose this fork only if you specifically need the VoVNet-oriented research setup or its historical behavior. For general Detectron2 work, upstream is the better default because this fork is old and heavily diverged.

Prefer upstream unless you specifically need this fork's training-loop and dataloader behavior. This fork looks stale and materially stripped down, so it is better as a legacy or specialized variant than as a general-purpose Detectron2 base.

Choose this fork if your priority is build, packaging, or CI convenience in a constrained environment. Choose upstream if you want the latest Detectron2 fixes, broader maintenance, and less merge debt.

Prefer this fork only if you need its specific project-level tweaks and can tolerate an old upstream base. If you want current Detectron2 fixes, deployment polish, and broader maintenance, upstream is the better choice.

Prefer this fork only if its DensePose and Windows-specific changes match an existing dependency. For most adopters, upstream Detectron2 is the better choice because this fork is stale, heavily diverged, and likely missing modern features and maintenance.