The collection of pre-trained AI models, and how they were converted, deployed. 中文
The ncnn framework enables cross-device deployment with the help of the vulkan api. We pre-train models via pytorch, tensorflow, paddle etc. and then convert them to ncnn models for final deployment on Windows, mac, linux, android, ios, WebAssembly and uni-app. However, model conversion is not a one-click process and needs to be handled manually. In order to extend the boundary applications of ncnn, we have created this repository to receive any cases of successful or failed conversions.
✅ : good to work ❌ : bad to work ⭕ : good to work, but not good to contribute 🤔 : not sure, but good to contribute 🔥🔥💥
We believe we will succeed in the end. 😂。
| Model | Year | Size | From | Type | Convert | NCNN | Hot |
|---|---|---|---|---|---|---|---|
| roop | 2023 | 276.7M | Onnx | face_swap | ✅ | 🤔 | 🔥 |
| nerf | 2023 | 0.1MB | Pytorch | instant-ngp | ✅ | ✅ | |
| codeformer | 2023 | 212.5M | Pytorch | face_restoration | ✅ | ✅ | 🔥 |
| vits | 2022 | 91MB | Pytorch | tts | ✅ | ✅ | 🔥 |
| stablediffusion | 2022 | 1.7GB | Pytorch | diffusion | ✅ | ✅ | 🔥 |
| sherpa | 2022 | 134MB | Pytorch | tts | ✅ | ✅ | 🔥 |
| DTLN | 2022 | 1.9MB | Pytorch | audio_denoising | ✅ | ✅ | |
| gpt2-chinese | 2022 | 357MB | nlp | ⭕ | |||
| MAT | 2022 | Pytorch | image_inpainting | ❌ | |||
| RVM | 2021 | 13.6MB | Pytorch | image_matting | ✅ | ||
| vitea | 2022 | 52.9MB | Pytorch | image_matting | ❌ | ||
| AnimeGanV3 | 2022 | Onnx | style_transfer | ❌ | |||
| HybridNets | 2022 | Pytorch | object_dection | ❌ | |||
| yolop | 2021 | Pytorch | object_dection | ❌ | 🤔 | 💥 | |
| pfld | 2019 | 4.9MB | Pytorch | face_dection | ❌ | ✅ | |
| Anime | 2021 | 18.8MB | Onnx | face_dection | ✅ | ⭕ | |
| CaiT | 2021 | 34.3MB | Pytorch | image_classification | ✅ | ||
| FastestDet | 2022 | 0.4MB | Pytorch | object_dection | ✅ | ✅ | 💥 |
| yolov7 | 2022 | 12.1MB | Pytorch | object_dection | ✅ | ✅ | |
| yolov6 | 2022 | 8.4MB | Pytorch | object_dection | ⭕ | ✅ | |
| yolov5 | 2021 | 2.3MB | Pytorch | object_dection | ✅ | ✅ | 💥 |
| yolo-fastestv2 | 2021 | 0.4MB | Pytorch | object_dection | ✅ | ✅ | 💥 |
| yolox | 2021 | 1.7MB | Pytorch | object_dection | ✅ | ✅ | |
| nanodet | 2020 | 2.3MB | Onnx | object_dection | ✅ | ✅ | |
| DenseNet | 2018 | 21.5MB | Pytorch | image_classification | ✅ | ✅ | |
| resnet18 | 2015 | 22.8MB | Pytorch | image_classification | ✅ | ✅ | |
| mobilenet_v2 | 2019 | 6.8MB | Pytorch | image_classification | ✅ | ✅ | |
| mobilenet_v3 | 2019 | 10.7MB | Pytorch | image_classification | ✅ | ✅ | |
| Res2Net | 2021 | 88.2MB | Pytorch | image_classification | ✅ | ✅ | |
| Res2Next50 | 2021 | 48.1MB | Pytorch | image_classification | ✅ | ✅ | |
| shufflenetv2 | 2018 | 4.4MB | Onnx | image_classification | ✅ | ✅ | |
| vgg16 | 2015 | 263MB | Pytorch | image_classification | ✅ | ✅ | |
| efficientnet | 2021 | 10.3MB | Pytorch | image_classification | ✅ | ✅ | |
| deeplabv3 | 2017 | 21.5MB | Pytorch | image_matting | ✅ | ✅ | |
| yolov7-mask | 2022 | 86.6MB | Pytorch | image_matting | 🤔 | ✅ | |
| deoldify | 2019 | 242MB | Onnx | image_inpainting | 🤔 | ✅ | |
| UltraFace | 2019 | 0.6MB | Pytorch | face_dection | ✅ | ✅ | |
| Anime2Real | 2022 | 22.2MB | Pytorch | style_transfer | ✅ | ✅ | |
| AnimeGanV2 | 2020 | 4.2MB | Pytorch | style_transfer | ✅ | ✅ | |
| styletransfer | 2016 | 3.2MB | Onnx | style_transfer | ✅ | ✅ | |
| ifrnet | 2022 | 5.6MB | Pytorch | video_frame_interpolation | ✅ | ||
| Rife | 2021 | 10MB | Onnx | video_frame_interpolation | ✅ | ||
| GFPGAN | 2021 | 214MB | Onnx | face_restoration | ✅ | 💥 |
see DeepFaceLive
waifu2x-ncnn-vulkan、realcugan-ncnn-vulkan、realEsrgan-ncnn-vulkan ...
see RealESRGAN
see MODNet
see BlazePose
see AnimeGanV2
- 824562395 【加群请备注你正在转换的新模型(2022 至今)】