- MegEngine_CU11: 包含 CUDA11 支持的 MegEngine Python Wheel 包
- MgeEditing：基于 MegEngine 的图像库
- Echo：一个优秀的算子库
- ClearML(Allegro Trains): ML/DL 开发和生产套件，包含 Experiment Manager / ML-Ops / Data-Management 三大核心模块
- MegEngine.js：MegEngine javascript 版本，可以在 javascript 环境中快速部署 MegEngine 模型
- commitlint config megengine：MegEngine commitlint 配置项
- Cascade RCNN ：一种 Cascade RCNN 实现
- Megvision：一些经典 CV 模型的实现和权重
- Models：MegEngine 实例训练代码，包含各类任务的基本实现，是初学者最好的参考物
- MgeConvert：MegEngine Traced module 模型格式转换器，支持 ONNX / TFLite / Caffe 等各类导出格式
- MegFile：一个可以完美抽象 S3、HTTP、本地文件等协议的 python 文件接口库，是 smart-open 库的升级版
- MegFlow：面向计算机视觉应用的流式计算框架，提供了一套可快速完成 AI 应用部署的视觉解析服务方案
- MegSpot：一款提供免费免登录、高效、专业、跨平台的图片&视频的对比的 PC 应用工具
- BaseCls：极其全面的分类模型库，提供了海量模型的训练代码和预训练权重，全部算法均可以快速部署到硬件上
- Swin Transformer：Swin Transformer 的实现，支持 DTR 功能减少显存占用量
- MegFLow-Cat_Feeder：基于
MegFlow框架的猫咪检测以及自动投喂的解决方案 - MegPeak：处理器性能测试工具
- Netron 已可视化 MegEngine TracedModule，欢迎使用 模型示例文件 体验
- MegCC：一个运行时超轻量，高效，移植简单的深度学习模型编译器
- BaseDet：提供了一些经典的检测 SOTA 模型以及相关组件
- mperf：一个算子性能调优工具箱
- MegEngine 原创论文复现：
- RepLKNet：[CVPR 2022] Scaling Up Your Kernels to 31x31: Revisiting Large Kernel Design in CNNs
- GyroFlow: [ICCV 2021] GyroFlow: Gyroscope-Guided Unsupervised Optical Flow Learning
- ICD: [NeurIPS 2021] Instance-Conditional Knowledge Distillation for Object Detection
- FINet: [AAAI 2022] FINet: Dual Branches Feature Interaction for Partial-to-Partial Point Cloud Registration
- OMNet: [ICCV 2021] OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration
- CREStereo: [CVPR 2022] Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlation
- D2C-SR: [ECCV 2022] D2C-SR: A Divergence to Convergence Approach for Real-World Image Super-Resolution
- FST-Matching: [ECCV 2022] Explaining Deepfake Detection by Analysing Image Matching
- zipfls: [ECCV 2022] Zipf's LS: Efficient One Pass Self-distillation with Zipf's Label Smoothing
- HDR-Transformer: [ECCV 2022] Ghost-free High Dynamic Range Imaging with Context-aware Transformer
- 经典论文模型对应结构的 MegEngine inference 函数：
- Masked Autoencoders Are Scalable Vision Learners
- Wide Residual Networks
- ResNeSt: Split-Attention Networks
- Visual Attention Network
- A ConvNet for the 2020s
- UniFormer: Unifying Convolution and Self-attention for Visual Recognition
- Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detection
- Probabilistic Anchor Assignment with IoU Prediction for Object Detection
- OTA: Optimal Transport Assignment for Object Detection
- Pyramid Scene Parsing Network
- GhostNet: More Features from Cheap Operations
- Squeeze-and-Excitation Networks
- OpenPose: Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields
- Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net
- HarDNet: A Low Memory Traffic Network
- Learning to Navigate for Fine-grained Classification
- FishNet: a versatile backbone for image, region, and pixel level prediction
- A Light CNN for Deep Face Representation with Noisy Labels
- Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
- Towards Compact Single Image Super-Resolution via Contrastive Self-distillation
- ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks
- Neighbor2Neighbor: Self-Supervised Denoising from Single Noisy Images
- Densely Connected Convolutional Networks
- Aggregated Residual Transformations for Deep Neural Networks
- Learning Transferable Visual Models From Natural Language Supervision
- Image Super-Resolution Using Very Deep Residual Channel Attention Networks
- Residual Dense Network for Image Super-Resolution
- BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation