登录

CoreML-Models

john-rocky活跃维护

把 PyTorch 模型转成 iOS/macOS 可直接用的 .mlpackage，每个模型都配转换脚本和 SwiftUI 示例，省去自己折腾 Core ML 的麻烦。

Core ML model zoo for iOS/macOS — PyTorch models converted to ready-to-use .mlpackage, each with a conversion script and SwiftUI sample app. Sibling repos cover Apple's Core AI framework (iOS/macOS 27) and on-device LLMs.

yoloe-v1发布于 2026-06-01 05:55 UTC1,895star +2 / 7 天178fork同步于 2026-11-10 11:46 UTC
Core ML端侧推理模型转换iOS 部署SwiftUI 示例
对比 GitHub 官网apple-siliconcore-aicoremlcoremltoolsdeep-learning

怎么上手

由 README 提炼，执行前请确认
原文 
README 未给出安装命令，使用方式是下载 .mlpackage 后打包进 Xcode 项目
硬件要求按 README 说明

README 未说明具体硬件要求；模型面向 iOS/macOS 的 Core ML 运行环境，未列出显存或 GPU 型号。

支持 Apple Silicon
下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼
没人工点评？写下你的经验

把 PyTorch 模型转成可直接用的 Core ML .mlpackage，省去自己转换和搭示例的功夫

把 PyTorch 模型转成 iOS/macOS 可直接用的 .mlpackage，每个模型都配转换脚本和 SwiftUI 示例，省去自己折腾 Core ML 的麻烦。

核心能力

  • 按图像分类、检测、分割、TTS 等任务分类整理模型目录
  • 每个模型提供 Google Drive 下载的 .mlpackage 文件
  • 部分模型附带转换脚本和 SwiftUI 示例工程
  • 覆盖 YOLO 系列、Stable Diffusion、SAM2 等热门模型
  • 标注各模型原始项目、数据集、体积与许可证

容易踩的坑

模型需从 Google Drive 手动下载，仓库本身不含权重文件；各模型许可证跟随原项目，商用前需逐个确认；.mlpackage 与 iOS 27 的 .aimodel 格式不通用，需另找 Core AI 仓库

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

最近发版

5 个版本yoloe-v1
全部 release
  1. yoloe-v1YOLOE — Open-Vocabulary Detection + Segmentation + Visual Prompts最新正式版2026-06-01 · 2026-06-01 05:55 UTC

    YOLOE converted to Core ML: real-time open-vocabulary detection + instance segmentation, with text and visual prompts. Sizes S (fast) and L (accurate).

    Region-embedding design: the detector is text-free and emits per-anchor region embeddings; the region-query similarity (BNContrastiveHead, folded into a 513-d augmented dot product) runs in Swift, so the vocabulary changes for free without re-running the image branch.

    • yoloe_detector_s/l — region-embedding detector + seg (image -> boxes, region_embeddings[1,513,8400], mask_coeffs, mask_protos)
    • reprta_s/l — RepRTA text-prompt refinement MLP (raw_tpe -> tpe)
    • visual_prompt_encoder_s/l — SAVPE visual prompt (image + 80x80 box mask -> vpe[1,1,512]); a drop-in for the text query
    • mobileclip_blt_text — Apple MobileCLIP B-LT text encoder (shared)
    • clip_vocab.json — CLIP BPE vocabulary (shared)

    Demo: sample_apps/YOLOEDemo (camera/photo/video + Visual tab). License: AGPL-3.0 (THU-MIG/yoloe); MobileCLIP is Apple's export.

  2. fastsam-v1FastSAM (YOLOv8-seg) CoreML2026-05-29 · 2026-05-29 07:05 UTC
    查看更新说明

    FastSAM instance segmentation (segment-everything) converted to Core ML, FP16. FastSAM-s (light, ~23 MB) and FastSAM-x (quality, ~138 MB). YOLOv8-seg architecture; outputs boxes + scores + mask coefficients + mask protos. License: AGPL-3.0 (Ultralytics). See sample_apps/FastSAMDemo and conversion_scripts/convert_fastsam.py.

  3. moge2-v1MoGe-2 ViT-B Normal (504×504, FP16)2026-04-08 · 2026-04-08 02:49 UTC
    查看更新说明

    MoGe-2 ViT-B + normal (CVPR 2025 Oral) — monocular 3D geometry estimation.

    Single CoreML mlpackage (~200 MB FP16, 504×504 fixed input). Predicts metric depth, surface normals, and confidence mask from a DINOv2 ViT-B/14 backbone.

    Variant: Ruicheng/moge-2-vitb-normal (104M params)

    OutputShapeDescription
    points(1, 504, 504, 3)3D point map (affine, exp-remapped)
    depth(1, 504, 504)Raw depth (multiply by metric_scale for meters)
    normal(1, 504, 504, 3)Surface normals, L2-normalized
    mask(1, 504, 504)Confidence mask (sigmoid, threshold at 0.5)
    metric_scale(1,)Scalar to convert raw depth → metric meters

    See sample_apps/MoGe2Demo/ for the iOS demo app.

  4. kokoro-v1Kokoro-82M CoreML v12026-04-07 · 2026-04-07 03:30 UTC
    查看更新说明

    First CoreML port of hexgrad/Kokoro-82M with on-device bilingual (English + Japanese) free-text input. Contains: Predictor (flexible 1-256 phonemes) + Decoder buckets (128/256/512 frames). FP32. Includes the mod-by-1 fix in iSTFTNet SineGen. See README.md → Text-to-Speech → Kokoro-82M for details.

  5. hypersd-v1Hyper-SD v1 — Single-Step SD1.5 Text-to-Image2026-04-06 · 2026-04-06 05:34 UTC
    查看更新说明

    4 CoreML models for Hyper-SD 1-step text-to-image (SD1.5 base + ByteDance Hyper-SD LoRA fused, 512×512 output, ~947 MB total).

    • HyperSDTextEncoder.mlpackage.zip (216 MB, FP16) — CLIP ViT-L text encoder
    • HyperSDUnetChunk1.mlpackage.zip (310 MB, 6-bit palettized) — Hyper-SD UNet first half (Split-Einsum attention)
    • HyperSDUnetChunk2.mlpackage.zip (290 MB, 6-bit palettized) — Hyper-SD UNet second half
    • HyperSDVAEDecoder.mlpackage.zip (87 MB, FP16) — Stable Diffusion VAE decoder

    Single-step generation via TCD scheduler. Runs on iPhone 15+ with Neural Engine.

    See conversion_scripts/convert_hypersd.py and sample_apps/HyperSDDemo.

用户点评

登录用户的真实使用记录，可按版本和硬件筛选
共 0 条

    还没有人点评。第一个说出真实体验的人最有价值。

    README 原文

    未翻译
    在 GitHub 打开

    CoreML-Models

    Maintained by Daisuke Majima — on-device AI for iOS & Android.

    Looking for Core AI on iOS 27 or macOS 27?

    For a .aimodel bundle or a custom local model in a new Swift app, the Core AI start page connects the model zoo, CoreAIKit, and downloadable demo apps. It includes the tested environment for each starting point. Coding agents can read the task index.

    This repository provides Core ML models and sample apps. Use the catalog below for existing Core ML integrations, earlier OS targets, or models available here; .mlpackage and .aimodel are different formats.

    How to use

    Take a look this model zoo, and if you found the CoreML model you want, download the model from google drive link and bundle it in your project. Or if the model have sample project link, try it and see how to use the model in the project. You are free to do or not.

    If you like this repository, please give me a star so I can do my best.

    Projects using these models

    • xLights — an open-source light show sequencer. On macOS, xLights 2026.18 downloads this repository's HTDemucs Core ML model on first use and splits a song's audio into four stems: drums, bass, other, and vocals (source).

    Section Link

    How to get the model

    You can get the model converted to CoreML format from the link of Google drive. See the section below for how to use it in Xcode. The license for each model conforms to the license for the original project.

    Image Classifier

    Efficientnet

    Google Drive LinkSizeDatasetOriginal ProjectLicense
    Efficientnetb022.7 MBImageNetTensorFlowHubApache2.0

    Efficientnetv2

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    Efficientnetv285.8 MBImageNetGoogle/autoMLApache2.02021

    VisionTransformer

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale.

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    VisionTransformer-B16347.5 MBImageNetgoogle-research/vision_transformerApache2.02021

    Conformer

    Local Features Coupling Global Representations for Visual Recognition.

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    Conformer-tiny-p1694.1 MBImageNetpengzhiliang/ConformerApache2.02021

    DeiT

    Data-efficient Image Transformers

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    DeiT-base384350.5 MBImageNetfacebookresearch/deitApache2.02021

    RepVGG

    Making VGG-style ConvNets Great Again

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    RepVGG-A033.3 MBImageNetDingXiaoH/RepVGGMIT2021

    RegNet

    Designing Network Design Spaces

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYear
    regnet_y_400mf16.5 MBImageNetTORCHVISION.MODELSMIT2020

    MobileViTv2

    CVNets: A library for training computer vision networks

    Google Drive LinkSizeDatasetOriginal ProjectLicenseYearConversion Script
    MobileViTv218.8 MBImageNetapple/ml-cvnetsapple2022Open In Colab

    Object Detection

    D-FINE

    Download LinkSizeOutputOriginal ProjectLicenseNoteSample Project
    dfine-n-coco13MBConfidence(MultiArray (Float32 300 × 80)), Coordinates (MultiArray (Float32 300 × 4))Peterande/D-FINEApache 2.0Input 640×640. Coordinates are normalized cxcywh. No NMS — filter by confidence threshold.peaceofcake DFINEDemo

    RF-DETR

    Download LinkSizeOutputOriginal ProjectLicenseNoteSample Project
    rfdetr-n-coco95MBConfidence(MultiArray (Float32 300 × 91)), Coordinates (MultiArray (Float32 300 × 4))roboflow/rf-detrApache 2.0Input 384×384. 91 classes (index 0 = background, 1-90 = COCO category IDs). Coordinates are normalized cxcywh. No NMS.peaceofcake DFINEDemo

    YOLOv5s

    Google Drive LinkSizeOutputOriginal ProjectLicenseNoteSample Project
    YOLOv5s29.3MBConfidence(MultiArray (Double 0 × 80)), Coordinates (MultiArray (Double 0 × 4))ultralytics/yolov5GNUNon Maximum Suppression has been added.CoreML-YOLOv5

    YOLOv7

    Google Drive LinkSizeOutputOriginal ProjectLicenseNoteSample ProjectConversion Script
    YOLOv7147.9MBConfidence(MultiArray (Double 0 × 80)), Coordinates (MultiArray (Double 0 × 4))WongKinYiu/yolov7GNUNon Maximum Suppression has been added.CoreML-YOLOv5Open In Colab

    YOLOv8

    Google Drive LinkSizeOutputOriginal ProjectLicenseNoteSample Project
    YOLOv8s45.1MBConfidence(MultiArray (Double 0 × 80)), Coordinates (MultiArray (Double 0 × 4))ultralytics/ultralyticsGNUNon Maximum Suppression has been added.CoreML-YOLOv5

    YOLOv9

    YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information. Uses PGI and GELAN architecture for efficient object detection.

    Download LinkSizeOutputOriginal ProjectLicenseYearNoteSample Project
    yolov9s.mlpackage.zip14 MBConfidence (MultiArray (Double 0 × 80)), Coordinates (MultiArray (Double 0 × 4))WongKinYiu/yolov9GPL-3.02024Non Maximum Suppression has been added.YOLOv9Demo

    YOLOv10

    YOLOv10: Real-Time End-to-End Object Detection. NMS-free architecture using consistent dual assignments — no post-processing needed.

    Download LinkSizeOutputOriginal ProjectLicenseYearNoteSample Project
    yolov10s.mlpackage.zip14 MBMultiArray (1 × 300 × 6)THU-MIG/yolov10AGPL-3.02024NMS-free end-to-end detection.YOLO26Demo

    YOLO11

    YOLO11: Ultralytics latest YOLO with improved backbone and neck architecture. 22% fewer parameters than YOLOv8 with higher mAP.

    Download LinkSizeOutputOriginal ProjectLicenseYearNoteSample Project
    yolo11s.mlpackage.zip18 MBConfidence (MultiArray (Double 0 × 80)), Coordinates (MultiArray (Double 0 × 4))ultralytics/ultralyticsAGPL-3.02024Non Maximum Suppression has been added.YOLOv9Demo

    YOLO26

    YOLO26: Edge-first vision AI with NMS-free end-to-end detection. Up to 43% faster CPU inference vs YOLO11 with DFL removal and ProgLoss.

    Download LinkSizeOutputOriginal ProjectLicenseYearNoteSample Project
    yolo26s.mlpackage.zip18 MBMultiArray (1 × 300 × 6)ultralytics/ultralyticsAGPL-3.02026NMS-free end-to-end detection.YOLO26Demo

    YOLO-World

    YOLO-World: Real-Time Open-Vocabulary Object Detection. Type any text query and detect it — no fixed class list. Uses CLIP text encoder for open-vocabulary matching.

    Download LinkSizeDescriptionOriginal ProjectLicenseYearSample Project
    yoloworld_detector.mlpackage.zip25 MBYOLO-World V2-S visual detectorAILab-CVC/YOLO-WorldGPL-3.02024YOLOWorldDemo
    clip_text_encoder.mlpackage.zip121 MBCLIP ViT-B/32 text encoderopenai/CLIPMIT2021—
    clip_vocab.json.zip1.6 MBBPE vocabulary for tokenizer————

    YOLOE

    YOLOE: Real-Time Open-Vocabulary Detection + Instance Segmentation. Detect and segment anything from a text query or a visual prompt (box an example object) — no fixed class list. Available in S (fast) and L (accurate). See YOLOEDemo for the region-embedding + MobileCLIP pipeline.

    Download LinkSizeDescriptionOriginal ProjectLicenseYearSample Project
    yoloe_detector_s.mlpackage.zip20 MBYOLOE-11s-seg region-embedding detector + segmentationTHU-MIG/yoloeAGPL-3.02025YOLOEDemo
    yoloe_detector_l.mlpackage.zip54 MBYOLOE-11l-seg region-embedding detector + segmentationTHU-MIG/yoloeAGPL-3.02025YOLOEDemo
    reprta_s.mlpackage.zip6 MBYOLOE RepRTA text-refinement MLP (S)THU-MIG/yoloeAGPL-3.02025—
    reprta_l.mlpackage.zip6 MBYOLOE RepRTA text-refinement MLP (L)THU-MIG/yoloeAGPL-3.02025—
    visual_prompt_encoder_s.mlpackage.zip20 MBYOLOE SAVPE visual-prompt encoder (S): image + box → queryTHU-MIG/yoloeAGPL-3.02025YOLOEDemo
    visual_prompt_encoder_l.mlpackage.zip54 MBYOLOE SAVPE visual-prompt encoder (L): image + box → queryTHU-MIG/yoloeAGPL-3.02025YOLOEDemo
    mobileclip_blt_text.mlpackage.zip121 MBApple MobileCLIP B-LT text encoder (shared)apple/ml-mobileclipApple2024—
    clip_vocab.json.zip1.6 MBBPE vocabulary for tokenizer (shared)————

    Multi-Object Tracking

    ByteTrack

    ByteTrack: Multi-Object Tracking by Associating Every Detection Box. Pure-Swift on-device tracker that adds persistent IDs on top of any detector above — an 8D Kalman filter plus two-stage IoU association, no appearance / ReID network.

    ImplementationSourcePaperLicenseYearNoteSample Project
    Pure Swift (no download)Tracker.swiftByteTrack (arXiv 2110.06864)MIT (this port) / Original20228D Kalman + two-stage IoU association, class-aware, greedy matching, lost-track buffer of 30 frames. Drop-in on top of any [Detection] stream.YOLO26Demo

    Segmentation

    U2Net

    Google Drive LinkSizeOutputOriginal ProjectLicense
    U2Net175.9 MBImage(GRAYSCALE 320 × 320)xuebinqin/U-2-NetApache
    U2Netp4.6 MBImage(GRAYSCALE 320 × 320)xuebinqin/U-2-NetApache

    IS-Net

    Google Drive LinkSizeOutputOriginal ProjectLicenseYearConversion Script
    IS-Net176.1 MBImage(GRAYSCALE 1024 × 1024)xuebinqin/DISApache2022Open In Colab
    IS-Net-General-Use176.1 MBImage(GRAYSCALE 1024 × 1024)xuebinqin/DISApache2022Open In Colab

    RMBG1.4

    RMBG1.4 - The IS-Net enhanced with our unique training scheme and proprietary dataset.

    Download LinkSizeOutputOriginal ProjectLicenseyearSample ProjectConversion Script
    RMBG_1_4.mlpackage.zip42 MB (INT8)Alpha mask 1024x1024briaai/RMBG-1.4Creative Commons2024RMBGDemoconvert_rmbg.py

    face-Parsing

    Google Drive LinkSizeOutputOriginal ProjectLicenseSample Project
    face-Parsing53.2 MBMultiArray(1 x 512 × 512)zllrunning/face-parsing.PyTorchMITCoreML-face-parsing

    Segformer

    Simple and Efficient Design for Semantic Segmentation with Transformers

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    SegFormer_mit-b0_1024x1024_cityscapes14.9 MBMultiArray(512 × 1024)NVlabs/SegFormerNVIDIA2021

    BiSeNetV2

    Bilateral Network with Guided Aggregation for Real-time Semantic Segmentation

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    BiSeNetV2_1024x1024_cityscapes12.8 MBMultiArrayycszen/BiSeNetApache2.02021

    DNL

    Disentangled Non-Local Neural Networks

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    dnl_r50-d8_512x512_80k_ade20k190.8 MBMultiArray[512x512]ADE20Kyinmh17/DNL-Semantic-SegmentationApache2.02020

    ISANet

    Interlaced Sparse Self-Attention for Semantic Segmentation

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    isanet_r50-d8_512x512_80k_ade20k141.5 MBMultiArray[512x512]ADE20Kopenseg-group/openseg.pytorchMITArXiv'2019/IJCV'2021

    FastFCN

    Rethinking Dilated Convolution in the Backbone for Semantic Segmentation

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    fastfcn_r50-d32_jpu_aspp_512x512_80k_ade20k326.2 MBMultiArray[512x512]ADE20Kwuhuikai/FastFCNMITArXiv'2019

    GCNet

    Non-local Networks Meet Squeeze-Excitation Networks and Beyond

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    gcnet_r50-d8_512x512_20k_voc12aug189 MBMultiArray[512x512]PascalVOCxvjiarui/GCNetApache License 2.0ICCVW'2019/TPAMI'2020

    DANet

    Dual Attention Network for Scene Segmentation(CVPR2019)

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    danet_r50-d8_512x1024_40k_cityscapes189.7 MBMultiArray[512x1024]CityScapesjunfu1115/DANetMITCVPR2019

    Semantic-FPN

    Panoptic Feature Pyramid Networks

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    fpn_r50_512x1024_80k_cityscapes108.6 MBMultiArray[512x1024]CityScapesfacebookresearch/detectron2Apache License 2.02019

    cloths_segmentation

    Code for binary segmentation of various cloths.

    Google Drive LinkSizeOutputDatasetOriginal ProjectLicenseyear
    clothSegmentation50.1 MBImage(GrayScale 640x960)fashion-2019-FGVC6facebookresearch/detectron2MIT2020

    easyportrait

    EasyPortrait - Face Parsing and Portrait Segmentation Dataset.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyearSwift sampleConversion Script
    easyportrait-segformer512-fp7.6 MBImage(GrayScale 512x512) * 9hukenovs/easyportraitCreative Commons2023easyportrait-coremlOpen In Colab

    MobileSAM

    Faster Segment Anything: Towards Lightweight SAM for Mobile Applications. MobileSAM replaces the heavy ViT-H image encoder with a lightweight ViT-Tiny encoder via decoupled knowledge distillation, making it ~60x smaller and ~40x faster than the original SAM.

    Download LinkSizeOutputOriginal ProjectLicenseYearSample Project
    MobileSAM.zip23 MB (Encoder 13 MB + Decoder 9.8 MB)Segmentation MaskChaoningZhang/MobileSAMApache 2.02023SamKit

    SAM2-Tiny

    SAM 2: Segment Anything in Images and Videos. SAM 2 extends promptable segmentation from images to videos using a streaming architecture with memory. The Tiny variant uses a Hiera-T backbone for efficient on-device inference.

    Download LinkSizeOutputOriginal ProjectLicenseYearSample Project
    SAM2Tiny.zip76 MB (ImageEncoder 64 MB + PromptEncoder 2 MB + MaskDecoder 9.8 MB)Segmentation Maskfacebookresearch/sam2Apache 2.02024SamKit

    FastSAM

    Fast Segment Anything — a YOLOv8-seg instance segmenter (not a SAM encoder/decoder): one forward pass segments everything and point/box prompts just select among them, the fastest SAM-family option for real-time use. FastSAM-s (light) / FastSAM-x (quality).

    Download LinkSizeOutputOriginal ProjectLicenseYearSample ProjectConversion Script
    FastSAM_s.mlpackage.zip~23 MB FP16Instance masksCASIA-IVA-Lab/FastSAMAGPL-3.02023FastSAMDemo · SamKitconvert_fastsam.py
    FastSAM_x.mlpackage.zip~138 MB FP16Instance masksCASIA-IVA-Lab/FastSAMAGPL-3.02023FastSAMDemo · SamKitconvert_fastsam.py

    Note: AGPL-3.0 (Ultralytics YOLOv8), unlike the Apache-2.0 SAM family.

    Video Matting

    MatAnyone

    pq-yang/MatAnyone (CVPR 2025) — temporally consistent video matting with object-level memory propagation. From a first-frame mask it tracks and refines an alpha matte across the whole clip, holding sharp edges (hair, semitransparent regions) far better than per-frame baselines.

    Download LinkSizeInputOutputOriginal ProjectLicenseYearSample ProjectConversion Script
    MatAnyone (5 mlpackages, ~111 MB FP16 total)111 MBimage [1,3,432,768] (per-frame state in Swift)alpha matte [1,1,432,768]pq-yang/MatAnyoneNTU S-Lab 1.02025MatAnyoneDemoconvert_matanyone.py

    See sample_apps/MatAnyoneDemo/README.md for the per-frame state machine, the 5-module split, and conversion details.

    Super Resolution

    Real ESRGAN

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    Real ESRGAN4x66.9 MBImage(RGB 2048x2048)xinntao/Real-ESRGANBSD 3-Clause License2021
    Real ESRGAN Anime4x66.9 MBImage(RGB 2048x2048)xinntao/Real-ESRGANBSD 3-Clause License2021

    GFPGAN

    Towards Real-World Blind Face Restoration with Generative Facial Prior

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    GFPGAN337.4 MBImage(RGB 512x512)TencentARC/GFPGANApache2.02021

    BSRGAN

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    BSRGAN66.9 MBImage(RGB 2048x2048)cszn/BSRGAN2021

    A-ESRGAN

    Google Drive LinkSizeOutputOriginal ProjectLicenseyearConversion Script
    A-ESRGAN63.8 MBImage(RGB 1024x1024)aesrgan/A-ESRGANNBSD 3-Clause License2021Open In Colab

    Beby-GAN

    Best-Buddy GANs for Highly Detailed Image Super-Resolution

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    Beby-GAN66.9 MBImage(RGB 2048x2048)dvlab-research/Simple-SRMIT2021

    RRDN

    The Residual in Residual Dense Network for image super-scaling.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    RRDN16.8 MBImage(RGB 2048x2048)idealo/image-super-resolutionApache2.02018

    Fast-SRGAN

    Fast-SRGAN.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    Fast-SRGAN628 KBImage(RGB 1024x1024)HasnainRaz/Fast-SRGANMIT2019

    ESRGAN

    Enhanced-SRGAN.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    ESRGAN66.9 MBImage(RGB 2048x2048)xinntao/ESRGANApache 2.02018

    UltraSharp

    Pretrained: 4xESRGAN

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    UltraSharp34 MBImage(RGB 1024x1024)Kim2019/CC-BY-NC-SA-4.02021

    SRGAN

    Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    SRGAN6.1 MBImage(RGB 2048x2048)dongheehand/SRGAN-PyTorch2017

    SRResNet

    Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    SRResNet6.1 MBImage(RGB 2048x2048)dongheehand/SRGAN-PyTorch2017

    LESRCNN

    Lightweight Image Super-Resolution with Enhanced CNN.

    Google Drive LinkSizeOutputOriginal ProjectLicenseyearConversion Script
    LESRCNN4.3 MBImage(RGB 512x512)hellloxiaotian/LESRCNN2020Open In Colab

    MMRealSR

    Metric Learning based Interactive Modulation for Real-World Super-Resolution

    Google Drive LinkSizeOutputOriginal ProjectLicenseyearConversion Script
    MMRealSRGAN104.6 MBImage(RGB 1024x1024)TencentARC/MM-RealSRBSD 3-Clause2022Open In Colab
    MMRealSRNet104.6 MBImage(RGB 1024x1024)TencentARC/MM-RealSRBSD 3-Clause2022Open In Colab

    DASR

    Pytorch implementation of "Unsupervised Degradation Representation Learning for Blind Super-Resolution", CVPR 2021

    Google Drive LinkSizeOutputOriginal ProjectLicenseyear
    DASR12.1 MBImage(RGB 1024x1024)The-Learning-And-Vision-Atelier-LAVA/DASRMIT2022

    SinSR

    wyf0912/SinSR — single-step diffusion-based super-resolution (CVPR 2024, ~113M params). Distilled from ResShift for one-step 4x upscaling. Uses a Swin Transformer UNet with VQ-VAE latent space.

    Left: bicubic 4x upscale, Right: SinSR single-step diffusion SR (128x128 → 512x512)

    3 CoreML models: VQ-VAE encoder, Swin-UNet denoiser (single step), and VQ-VAE decoder with vector quantization.

    Download LinkSizeInputOutputOriginal ProjectLicenseYearSample ProjectConversion Script
    SinSR_Encoder.mlpackage.zip39 MBimage [1,3,1024,1024]latent [1,3,256,256]wyf0912/SinSRS-Lab2024SinSRDemoconvert_sinsr.py
    SinSR_Denoiser.mlpackage.zip420 MBinput [1,6,256,256]predicted_latent [1,3,256,256]
    SinSR_Decoder.mlpackage.zip58 MBlatent [1,3,256,256]image [1,3,1024,1024]

    See sample_apps/SinSRDemo/README.md for the inference pipeline and conversion details.

    Low Light Enhancement

    StableLLVE

    Learning Temporal Consistency for Low Light Video Enhancement from Single Images.

    Google Drive LinkSizeOutputOriginal ProjectLicenseYear
    StableLLVE17.3 MBImage(RGB 512x512)zkawfanx/StableLLVEMIT2021

    Zero-DCE

    Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement

    Google Drive LinkSizeOutputOriginal ProjectLicenseYearConversion Script
    Zero-DCE320KBImage(RGB 512x512)Li-Chongyi/Zero-DCESee Repo2021Open In Colab

    Retinexformer

    Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancement

    <img width="256" src="https://github.com/john-rocky/PersonSegmentationSampler/assets/2327

    相关项目

    同分类里 star 最高的项目
    对比全部 →
    ultralytics
    ultralytics·Python

    YOLO 系列官方实现库，一套 PyTorch 代码覆盖目标检测、实例与语义分割、图像分类、姿态估计和目标跟踪，训练到部署全流程打通。

    目标检测+4检测 / 分割
    63270+8712kv8.4.2082026-11-10 09:31 UTC
    yolov5
    ultralytics·Python

    基于 PyTorch 的目标检测全家桶，覆盖训练、推理、实例分割、图像分类和模型导出，开箱即用，适合快速落地视觉任务。

    目标检测+4检测 / 分割
    58258+517kv7.02026-11-07 16:59 UTC
    pytorch-image-models
    huggingface·Python

    收录最全的 PyTorch 图像骨干网络合集，涵盖 ResNet、ViT、ConvNeXt 等主流模型，自带训练、评估、推理和导出脚本及预训练权重，省去自己搭网络和找权重的麻烦。

    图像分类+4检测 / 分割
    37272+75.2kv1.0.312026-11-07 23:22 UTC
    detectron2
    facebookresearch·Python

    Detectron2 是 Facebook 开源的目标检测与分割平台，提供模块化架构和丰富预训练模型，方便快速搭建和训练视觉识别任务。

    目标检测+4检测 / 分割
    34980+57.9kv0.62026-10-27 09:21 UTC