/jump to repo

Repos Recent Bookmarks Watched Notes Tags Discover Compare Stats GitHub

Repository brief

roboflow/supervision

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 2026-03-30T15:56:59.148Z

1mo ago

roboflow/supervision

roboflow/supervision is an actively maintained Python computer-vision utilities library for reusable detection, visualization, and dataset tooling. It is stable, widely adopted, and still evolving, with a very large fork and star count and recent commits on the default `develop` branch.

Loading tags...

Stars36,797

Forks3,157

Default branchdevelop

Last pushed2026-03-30T15:37:23Z

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

Maintenance:

Magnitude:

Sort:

Selected

Prefer this fork only if you need its custom image and API changes and are willing to own divergence. If you want a stable, drop-in supervision library with the latest upstream behavior, upstream is the safer choice.

Prefer this fork only if you specifically need its custom annotation/detection changes and are comfortable maintaining a long-lived divergence. If you want a stable, broadly compatible Supervision base, upstream is the better choice.

Choose the fork only if you need its local fixes or package-layout changes and are prepared to own the maintenance burden. If you want current Supervision capabilities and lower upgrade risk, upstream is the better default.

Choose upstream unless you specifically need this fork’s custom detection/metrics/tracker behavior and are prepared to maintain a large, stale divergence yourself. For most adopters, upstream is the safer and lower-risk choice.

Choose this fork only if its local annotator/crop workflow changes are the point. Otherwise, upstream is the better default because this fork is stale, materially behind, and likely missing newer fixes and features.

Prefer the upstream project unless you specifically need the fork's custom detection/VLM work or package-layout changes and are prepared to own maintenance. For most adopters, the fork looks too stale and too far behind upstream to be the safer default.

Choose this fork if you want the added image/color utilities and are comfortable owning a materially diverged codebase. Choose upstream if compatibility, documentation completeness, and easy upgrades matter more.

Choose upstream unless you specifically need this branch's customizations and are prepared to own the maintenance burden. This fork looks like a substantial, stale divergence rather than a drop-in improvement, so it favors maintainers of a bespoke variant over new adopters.

Prefer upstream unless you specifically need this fork's older customized behavior. Choose this fork only if you are deliberately staying on a legacy snapshot and can absorb the maintenance burden of a large divergence.

Choose upstream unless you specifically need this fork's older, customized behavior. This fork is materially behind and looks better as a legacy/private variant than as a starting point for new work.

Fork comparison

alenque-82/supervision

38/100

stale

significant_divergence

Prefer this fork only if you need its custom image and API changes and are willing to own divergence. If you want a stable, drop-in supervision library with the latest upstream behavior, upstream is the safer choice.

Likely purpose

To serve as a long-lived fork for experimenting with and carrying custom computer-vision utility changes, especially around image manipulation, annotator behavior, and API cleanup.

Best for

Teams intentionally maintaining a custom CV utility stack.; Developers who want image helpers and annotator behavior tailored to their own pipeline.; Projects willing to trade upstream compatibility for local API control.; Users who can absorb fork-specific maintenance and do not need a drop-in upstream replacement.

Additional features

Added `tint_image` and `grayscale_image` image-processing helpers.
Added support work around visualizing examples and image-utility documentation.
Reworked package layout and internal naming conventions, suggesting a broader API modernization effort.
Expanded and refactored detection, annotator, and metrics internals rather than just patching a single subsystem.
Expands API, schema, or SDK surface area for downstream integrations and generated clients.
Changes agent workflow behavior, such as prompt handling, handoffs, worktrees, memory, or orchestration flow.

Missing features

Dropped `create tiles` from the public API.
Deprecated older `overlay_image` image behavior, which may break existing image-compositing workflows.
Lags upstream by 219 commits, so newer upstream fixes and features are not present here yet.
The large refactor and package move may make some upstream paths, imports, or tests harder to reuse directly.
Some upstream behavior appears to be removed, disabled, or intentionally stripped down in this fork.
It trails upstream by 219 commits, so some recent upstream features and fixes are likely not present yet.

Strengths

More control over image augmentation and annotation workflows.
Potentially cleaner or more modern internal API shape.
Broad refactor indicates the fork is being actively adapted for a specific downstream use case.
May be useful if you need custom visualization primitives not aligned with upstream defaults.

Risks

High divergence from upstream increases merge and maintenance cost.
Removed or deprecated public APIs can break existing user code.
Staleness relative to upstream means missed bug fixes and feature additions.
Package/layout changes may complicate upgrades or dependency management.