timm (pytorch-image-models)

Checkpoints remapped from official repository at https://github.com/DingXiaoH/RepVGG

Checkpoints remapped from official repo at https://github.com/idstcv/GPU-Efficient-Networks

Converted to PyTorch from https://github.com/google-research/vision_transformer

Aug 12, 2020

• New/updated weights from training experiments
  • EfficientNet-B3 - 82.1 top-1 (vs 81.6 for official with AA and 81.9 for AdvProp)
  • RegNetY-3.2GF - 82.0 top-1 (78.9 from official ver)
  • CSPResNet50 - 79.6 top-1 (76.6 from official ver)
• Add CutMix integrated w/ Mixup. See pull request for some usage examples
• Some fixes for using pretrained weights with in_chans != 3 on several models.

Aug 5, 2020

Universal feature extraction, new models, new weights, new test sets.

• All models support the features_only=True argument for create_model call to return a network that extracts feature maps from the deepest layer at each stride.
• New models
  • CSPResNet, CSPResNeXt, CSPDarkNet, DarkNet
  • ReXNet
  • (Modified Aligned) Xception41/65/71 (a proper port of TF models)
• New trained weights
  • SEResNet50 - 80.3 top-1
  • CSPDarkNet53 - 80.1 top-1
  • CSPResNeXt50 - 80.0 top-1
  • DPN68b - 79.2 top-1
  • EfficientNet-Lite0 (non-TF ver) - 75.5 (submitted by @hal-314)
• Add 'real' labels for ImageNet and ImageNet-Renditions test set, see results/README.md
• Test set ranking/top-n diff script by @KushajveerSingh
• Train script and loader/transform tweaks to punch through more aug arguments
• README and documentation overhaul. See initial (WIP) documentation at https://rwightman.github.io/pytorch-image-models/
• adamp and sgdp optimizers added by @hellbell

ReXNet weights from https://github.com/clovaai/rexnet#pretrained remapped for timm model changes

These are a mirror of weights from the official repository (https://github.com/zhanghang1989/ResNeSt ) to avoid issues with hosting changes/relocation

RegNet weights cleaned and remapped from https://github.com/facebookresearch/pycls/blob/master/MODEL_ZOO.md

Changes:

• first layer remapped from BGR to RGB
• cleaned out training details such as optimizer state, etc and leave just model state_dict (1/2 size)
• map layer names to mine

Weights copied and cleaned (just state dict) from https://github.com/mrT23/TResNet/blob/master/MODEL_ZOO.md and other MIIL weight releases hosted at (*.aliyuncs.com) for more consistent/fast transfer speeds and avoidance of downtime.

These weights are re-hosted from original repository (https://github.com/mehtadushy/SelecSLS-Pytorch) with permission of the author, Dushyant Mehta (@mehtadushy), under a CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/legalcode) license.

SelecSLS (core) Network Architecture as proposed in "XNect: Real-time Multi-person 3D Human Pose Estimation with a Single RGB Camera, Mehta et al." https://arxiv.org/abs/1907.00837

HRNet weights downloaded from official impl OneDrive links at: https://github.com/HRNet/HRNet-Image-Classification. Rehosted here with SHA hash for hub/modelzoo download compatibility.

Res2Net weights from https://github.com/gasvn/Res2Net for easier/faster access from North America that's compatible with model_zoo load_url

https://pypi.org/project/timm/

These weights have all originated from Cadene's Pretrained model repository: https://github.com/Cadene/pretrained-models.pytorch

I'm re-hosting some of the weights here that I use more often to reduce download times as the US/Canada to France link can be slow.

All weights present here were either trained by me with the code in this repository or ported by me from original implementations.