object detection / instance segmentation
PyTorch-based toolbox for object detection and instance segmentation with a rich set of models and training utilities.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Ultra-lightweight anchor-free detector, choose for mobile/edge deployment.
NanoDet-Plus⚡Super fast and lightweight anchor-free object detection model. 🔥Only 980 KB(int8) / 1.8MB (fp16) and run 97FPS on cellphone🔥
General CV training library including detection, choose for multi-task needs.
CVNets: A library for training computer vision networks
Simpler PyTorch detection framework, choose for lightweight experimentation.
一个简单方便的目标检测框架(PyTorch环境可直接运行，不需要cuda编译)，支持Faster_RCNN、Cascade_RCNN、Yolo系列、SSD等经典网络。
PaddlePaddle detection toolkit, choose if using Paddle ecosystem.
Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.
DETR-focused detection platform, choose for transformer-based research.
detrex is a research platform for DETR-based object detection, segmentation, pose estimation and other visual recognition tasks.
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Standalone Mask R-CNN, choose if you only need that model.
Mask R-CNN for object detection and instance segmentation on Pytorch
Mask R-CNN in Keras/TensorFlow, choose for TF workflows.
Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
Legacy Caffe Mask R-CNN, only for Caffe compatibility.
Implementation of Mask-RCNN in Caffe https://arxiv.org/pdf/1703.06870.pdf
Older Mask R-CNN, choose for a minimal implementation, but stale.
Standalone Faster R-CNN, choose if you only need that model.
Faster RCNN with PyTorch
Older Faster R-CNN implementation, only for legacy use.
pytorch based implementation faster rcnn
Specific YOLO variant with TensorRT, choose for that exact model.
🔥🔥🔥🔥 (Earlier YOLOv7 not official one) YOLO with Transformers and Instance Segmentation, with TensorRT acceleration! 🔥🔥🔥
Faster R-CNN in TensorFlow, choose for TF-based projects.
Faster-RCNN in Tensorflow
Cascade R-CNN in Detectron, legacy implementation.
Cascade R-CNN in Detectron
Boundary-preserving Mask R-CNN variant, choose for that specific model.
[ECCV 2020] Boundary-preserving Mask R-CNN
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
iOS example app for TF detection, not a swappable library.
An iOS application of Tensorflow Object Detection with different models: SSD with Mobilenet, SSD with InceptionV2, Faster-RCNN-resnet101
These projects were analysed and named mmdetection among their alternatives. The relationship is not symmetric — how mmdetection rates them is a separate judgement, made when mmdetection is analysed in its own right.
calls mmdetection “Use for production object detection with a large set of ready-made detector architectures.”
deep learning for image processing including classification and object-detection etc.
calls mmdetection “PyTorch toolbox with huge model zoo; choose for model variety and community support, but requires PyTorch.”
Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
calls mmdetection “choose for a config-driven object-detection toolbox with many models; note it has not been updated in about 2 years.”
State-of-the-Art Deep Learning scripts organized by models - easy to train and deploy with reproducible accuracy and performance on enterprise-grade infrastructure.
calls mmdetection “Mature PyTorch toolbox with a large model zoo — choose for PyTorch-centric workflows, though less actively maintained now.”
Object Detection toolkit based on PaddlePaddle. It supports object detection, instance segmentation, multiple object tracking and real-time multi-person keypoint detection.
calls mmdetection “Choose if you want a config-driven toolbox of many detector architectures; note it has had no pushes for two years.”
Implementation of paper - YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors
calls mmdetection “Use it if you need a modular, versatile detection toolbox with many state-of-the-art architectures.”
YOLOv10: Real-Time End-to-End Object Detection [NeurIPS 2024]
calls mmdetection “choose it when you want a broad OpenMMLab toolbox with many configurable detection/segmentation models rather than a single model package.”
RF-DETR is a real-time object detection and segmentation model architecture developed by Roboflow, SOTA on COCO, designed for fine-tuning. [ICLR 2026]
calls mmdetection “Choose when you want a modular benchmark toolbox with many detector implementations and reproducible configs, using the MMDetection ecosystem rather than CenterNet's single model.”
Object detection, 3D detection, and pose estimation using center point detection:
calls mmdetection “Choose when you need a modular detection/segmentation toolbox supporting many models rather than a single-algorithm repo.”
SOLO and SOLOv2 for instance segmentation, ECCV 2020 & NeurIPS 2020.
calls mmdetection “Choose when you need a maintained PyTorch toolbox with many detectors/segmenters and config-driven training, rather than a single Mask R-CNN repo.”
Mask R-CNN for object detection and instance segmentation on Pytorch
calls mmdetection “Use it for a config-driven closed-set detection zoo, but it has gone unmaintained for two years and has no open-vocabulary weights.”
[CVPR 2024] Real-Time Open-Vocabulary Object Detection