timm
PyTorch collection of image models and backbones with pretrained weights and training and export scripts.
About
Nearly every modern computer vision baseline lives in timm, Ross Wightman's collection of PyTorch image models that covers more than 100 architecture families and over 1,000 pretrained weight configurations, from ResNet, EfficientNet, and ConvNeXt to ViT, Swin, MobileNetV4, and recent additions such as DINOv3 and SigLIP-2 backbones. Beyond model definitions, the library ships reference training, validation, and inference scripts, a long list of optimizers running from AdamW and LAMB to Muon, Kron, and MARS, augmentation strategies including Mixup, CutMix, and RandAugment, learning rate schedulers, and consistent feature extraction APIs that make any backbone usable in detection or segmentation pipelines. A single pip install timm and a call to timm.create_model with pretrained weights pulled from the Hugging Face Hub is enough to start, on CPU or GPU. The project is Apache 2.0 licensed, hosted under the Hugging Face organization, and past 37,000 GitHub stars, with version 1.0.28 shipping in July 2026. It is the de facto backbone library for computer vision research and production transfer learning alike.
Should you use timm?
Pick it when
Pick timm when you need a pretrained image backbone to fine-tune for classification or plug into a detection or segmentation model, with consistent feature extraction across ResNet, ConvNeXt, ViT, and newer families from one API.
Look elsewhere when
Skip it if you need a complete detector or segmenter with heads, losses, and task training loops, where MMDetection, Detectron2, or Ultralytics YOLO fit, or zero-shot labels from text, where CLIP or SigLIP fit better.
Alternatives to timm
- DINOv2
Frozen self-supervised features that work with a linear head when labels are scarce, a much narrower model family than timm's 1,000-plus weights.
- MMDetection
Supplies full detection heads, necks, and training pipelines; timm only gives you the backbone to put inside them.
- Transformers
Broader Hugging Face library spanning vision, text, and audio models, better for multimodal work, while timm goes deeper on image backbones and training recipes.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Price
- Free
- Platform
- Local/Desktop
- Difficulty
- Easy (2/5)
- License
- Apache-2.0
- Added
- Jul 29, 2026