https://github.com/d-li14/regnet.pytorch

PyTorch-style and human-readable RegNet with a spectrum of pre-trained models
https://github.com/d-li14/regnet.pytorch

imagenet neural-architecture-search pretrained-models pytorch regnet resnext

Last synced: 1 day ago
JSON representation

PyTorch-style and human-readable RegNet with a spectrum of pre-trained models

• Host: GitHub
• URL: https://github.com/d-li14/regnet.pytorch
• Owner: d-li14
• License: mit
• Created: 2020-06-06T15:35:20.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2021-03-19T16:23:57.000Z (over 3 years ago)
• Last Synced: 2024-08-04T01:15:46.648Z (3 months ago)
• Topics: imagenet, neural-architecture-search, pretrained-models, pytorch, regnet, resnext
• Language: Python
• Homepage: https://arxiv.org/abs/2003.13678
• Size: 16.6 KB
• Stars: 68
• Watchers: 6
• Forks: 13
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-image-classification - unofficial-pytorch: https://github.com/d-li14/regnet.pytorch

README

        
# RegNet Implementation with TorchVision Style

PyTorch implementation of [Designing Network Design Spaces](https://arxiv.org/abs/2003.13678) by Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, and Piotr Dollár.

Compared to the [official codebase](https://github.com/facebookresearch/pycls), this repository follows the [torchvision's ResNeXt](https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py) style, which is expected to be more easily interpreted and utilized by pre-existing downstream applications.

We train the following models on 8x TITAN XP GPUs with 12G VRAM. During the first five epochs, we linearly ramp up the learning rate from 0.1.

# Pre-trained Models

| Model                                                        | Params (M) | GFLOPs | Batch size | Top-1 acc (%) (our impl.) | Top-1 acc (%) (official) |

| ------------------------------------------------------------ | ---------- | ------ | ---------- | ------------------------- | ------------------------ |

| [RegNetX-200M](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EcDivRLCy7BHmVuoKVyqfZsB8t7OpFoCEdnLOD495UKWCw?e=r0h5fh) | 2.685      | 0.199  | 1024       | 68.210                    | 68.9                     |

| [RegNetX-400M](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EaL_0Di7OK5DsCLtvGcw418BqZGg5BD875kOIFMnALcMLQ?e=1mEB0v) | 5.158      | 0.398  | 1024       | 72.278                    | 72.7                     |

| [RegNetX-600M](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/ETFwqAcWWctLh3dPaCm0R5YB4xJVvoGdCTuwYYQzJiq35g?e=8OeH7k) | 6.196      | 0.601  | 1024       | 73.862                    | 74.1                     |

| [RegNetX-800M](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/Ecd7nKqHLnZCmlgitigejWIBYLhcpqkDCoBx_CEILtQcCg?e=8Xt961) | 7.260      | 0.800  | 1024       | 74.940                    | 75.2                     |

| [RegNetX-1.6G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EXOeBD6xco5JmvLziY4zySEB1bR00A7DqCx9t4IbI_MAng?e=ZG5PxS) | 9.190      | 1.603  | 1024       | 76.706                    | 77.0                     |

| [RegNetX-3.2G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EQ1o8qVNLuhBg21Kgf_bss8BHFrhm8PLI3xMrMtD7a192Q?e=RG2LoH) | 15.296     | 3.177  | 512        | 78.188                    | 78.3                     |

| [RegNetX-4.0G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/ET7rz66druZGqPe-IFC21MQBd_kcLoYwXIoR9YQbJpGOqA?e=wfYSsA) | 22.118     | 3.965  | 512        | 78.690                    | 78.6                     |

| [RegNetX-6.4G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EQkGuWBHehlDnkr3cASqgS4Btul3Lb_iuO4IGHIeHrkWbA?e=ndLLQs) | 26.209     | 6.460  | 512        | 79.152                    | 79.2                     |

| [RgeNetX-8.0G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EYTpeCq4OnNIr9ly3KmokywBodWSZHHBNPhiwirhk9Urag?e=PDsrFu) | 39.573     | 7.995  | 512        | 79.380                    | 79.3                     |

| [RegNetX-12G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EWEqa2PJVdxOj-M95XlLFVIBns9cnbdV6V6ASl-lyHzwyw?e=XVhG10) | 46.106     | 12.087 | 256        | 79.998                    | 79.7                     |

| [RegNetX-16G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/ETjNce9S9bxGgU09rLXoXucBBNLbo3t8zdtyriK-Vc8Eww?e=CrNU6u) | 54.279     | 15.941 | 256        | 80.118                    | 80.0                     |

| [RegNetX-32G](https://hkustconnect-my.sharepoint.com/:u:/g/personal/dlibh_connect_ust_hk/EReWI0v2kvVBpKAGWRy2Hb0BKaIk6wx-VbkFBqYoE-YQZw?e=dT0dos) | 107.812    | 31.736 | 256        | 80.516                    | 80.5                     |

# Citation

```bibtex

@InProceedings{Radosavovic_2020_CVPR,

author = {Radosavovic, Ilija and Kosaraju, Raj Prateek and Girshick, Ross and He, Kaiming and Doll{\'a}r, Piotr},

title = {Designing Network Design Spaces},

booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},

month = {June},

year = {2020}

}

```