msracver/Relation-Networks-for-Object-Detection

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1104 stars · 187 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Official MXNet implementation of the CVPR 2018 paper 'Relation Networks for Object Detection', which adds an attention-based relation module to Faster R-CNN to replace NMS and improve detection accuracy. It includes training and evaluation code on the COCO dataset with ResNet-101 backbones.

Use cases

• reproduce relation network object detection results on COCO
• train Faster R-CNN with a learned relation module instead of NMS
• study attention-based duplicate removal for object detection
• compare learn NMS versus soft-NMS baselines
• build object detection research on top of a published CVPR 2018 model

When to choose

• you need the official reference implementation of the Relation Networks paper
• you work with MXNet and want a research-grade object detection codebase
• you want to experiment with learned NMS and relation modules in Faster R-CNN

When to avoid

• you need a production object detection system with active maintenance
• you prefer PyTorch or TensorFlow tooling
• you need support for recent CUDA versions or modern GPU hardware

Facets

library · maturity maintenance

machine-learning deep-learning computer-vision image-processing computer-vision deep-learning machine-learning python object-detection mxnet relation-networks faster-rcnn nms research-code cvpr-2018 linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="msracver/Relation-Networks-for-Object-Detection")