msracver/Deformable-ConvNets

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

4121 stars · 948 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Official MXNet implementation of Deformable Convolutional Networks (ICCV 2017) and R-FCN, including deformable convolution and ROI pooling operators. It provides training and testing code for object detection and segmentation models such as Deformable FPN and Deformable R-FCN.

Use cases

• implement deformable convolution layers in detection models
• reproduce COCO object detection results with Deformable FPN
• train Deformable R-FCN with Soft NMS
• run Deformable ConvNets v2 operators
• research learnable spatial sampling in CNNs

When to choose

• you need the official reference implementation of deformable convolutions on MXNet
• you want to reproduce the Deformable ConvNets or R-FCN papers
• you need the corrected DCNv2 operator with boundary handling

When to avoid

• you want a maintained PyTorch implementation - use mmdetection instead
• you need a production-ready detection framework
• you don't work with MXNet or legacy CUDA builds

Facets

library · maturity maintenance

deep-learning computer-vision image-processing computer-vision deep-learning machine-learning python windows deformable-convolution object-detection mxnet rfcn instance-segmentation research-code linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="msracver/Deformable-ConvNets")