jytime/Mask_RCNN_Pytorch Alternatives

Choose if you want PANet's path-aggregation architecture for potentially better instance segmentation, accepting 7.5-year-old unmaintained code.

PANet for Instance Segmentation and Object Detection

Choose when you need Mask Scoring R-CNN's mask-IoU scoring head for more reliable mask confidences than vanilla Mask R-CNN.

Codes for paper "Mask Scoring R-CNN".

Choose for a unified transformer-based detector/segmenter if you want a modern architecture and Apache-2.0 licensing, not a drop-in Mask R-CNN replacement.

[CVPR 2023] Official implementation of the paper "Mask DINO: Towards A Unified Transformer-based Framework for Object Detection and Segmentation"

Choose when you need a maintained PyTorch toolbox with many detectors/segmenters and config-driven training, rather than a single Mask R-CNN repo.

OpenMMLab Detection Toolbox and Benchmark

Choose when you want anchor-free, direct instance segmentation (SOLOv2) instead of region-based Mask R-CNN.

SOLO and SOLOv2 for instance segmentation, ECCV 2020 & NeurIPS 2020.

Choose if your stack is MXNet and you need a framework covering detection and instance recognition, not PyTorch.

A Simple and Versatile Framework for Object Detection and Instance Recognition

Choose if you are standardized on PaddlePaddle and need a current detection/segmentation toolkit with deployment tools.

Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.

Choose if your stack is TensorFlow and you need a Mask R-CNN implementation there, rather than the seed's PyTorch version.

Mask RCNN in TensorFlow

Choose when you want a framework-agnostic detection/segmentation pipeline over multiple backends, rather than a dedicated Mask R-CNN repo.

An Agnostic Computer Vision Framework - Pluggable to any Training Library: Fastai, Pytorch-Lightning with more to come

Choose for universal image segmentation with a transformer architecture if you want panoptic/semantic/instance masks in one model, not Mask R-CNN.

Code release for "Masked-attention Mask Transformer for Universal Image Segmentation"

Choose if you are locked into MXNet and need Mask R-CNN there, not PyTorch.

An MXNet implementation of Mask R-CNN

Choose when you need real-time instance segmentation with a single-stage fully convolutional model, rather than Mask R-CNN's two-stage accuracy.

A simple, fully convolutional model for real-time instance segmentation.

Choose only for a classic Fast R-CNN detection baseline without instance masks, on an older Caffe/MATLAB stack.

Fast R-CNN

Choose for FAIR's modular PyTorch Mask R-CNN if you can accept an archived, 3.6-year-old codebase.

Fast, modular reference implementation of Instance Segmentation and Object Detection algorithms in PyTorch.

Choose only if you need the original Faster R-CNN Matlab code for detection without instance masks.

Faster R-CNN

Choose for 2D/3D medical imaging detection and segmentation if you need clinical-domain adaptations, but it is narrower than the seed.

The Medical Detection Toolkit contains 2D + 3D implementations of prevalent object detectors such as Mask R-CNN, Retina Net, Retina U-Net, as well as a training and inference framework focused on dealing with medical images.

Choose if you only need PyTorch Faster R-CNN detection, without instance masks.

A faster pytorch implementation of faster r-cnn

Choose for FAIR's Caffe2-based detection platform if you need legacy Detectron models, but it is archived and superseded by Detectron2.

FAIR's research platform for object detection research, implementing popular algorithms like Mask R-CNN and RetinaNet.

Choose if you need a TensorFlow Faster R-CNN detection baseline without masks.

Tensorflow Faster RCNN for Object Detection

Choose as a minimal PyTorch Faster R-CNN learning example, but it lacks instance segmentation.

A simplified implemention of Faster R-CNN that replicate performance from origin paper