# CharlesShang/DCNv2

Deformable Convolutional Networks v2 with Pytorch

Repository: https://github.com/CharlesShang/DCNv2
Canonical: https://ross.abutalabs.com/products/dcnv2
Language: C++
License: BSD-3-Clause
License Family: permissive
Topics: pytorch, deformable-convolutional-networks, cnn
Last push: 2026-06-11T04:46:21+00:00

## Health v2 (maintenance only)
Score: 71/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 87, release rhythm 35, longevity 100
- inputs: {"age_days": 2828, "days_push": 83, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1323, forks 395 (observed 2026-08-28T04:04:22.136236+00:00)

## What it is
A PyTorch implementation of Deformable Convolutional Networks v2 (DCNv2), providing CUDA/CPU operators for deformable convolution as a drop-in layer. It is an adaptation of the official MXNet Deformable-ConvNets operator for PyTorch 1.x.

## Use cases
- add deformable convolution layers to a PyTorch model
- implement DCNv2 for object detection backbones
- build instance or semantic segmentation models with deformable convolutions
- run gradient checks for custom CUDA convolution operators
- port the official Deformable-ConvNets operator to PyTorch 1.x

## When to choose
- you need deformable convolution in PyTorch with GPU acceleration
- you want a battle-tested DCNv2 operator with passing gradient checks
- you are reproducing detection/segmentation papers that rely on DCNv2

## When to avoid
- you use PyTorch 2.x or newer where torchvision ships torchvision.ops.deform_conv2d
- you need a maintained library with active bug fixes
- you need deformable attention (DCNv3/v4) rather than deformable convolution

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing
- domain: deep-learning, computer-vision, machine-learning
- platform: python, cpp, windows
- tags: pytorch, deformable-convolution, cnn, cuda-extension, object-detection, gpu, linux, macos

## Member repositories
- CharlesShang/DCNv2 (main) score 71

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:22.136236+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:47:02.935823+00:00, confidence not recorded.
  - readme: https://github.com/CharlesShang/DCNv2 (fetched 2026-08-28T04:04:22.136236+00:00, sha 0142136e4f12)
- Data as of 2026-08-30T08:39:29.467469+00:00.