kmodels 0.2.5

Pretrained keras 3 vision models

Keras Models 🚀

📖 Introduction

Keras Models (kmodels) is a collection of models with pretrained weights, built entirely with Keras 3. It supports a range of tasks, including classification, object detection (DETR, RT-DETR, RT-DETRv2, RF-DETR, D-FINE), segmentation (SAM, SAM2, SAM3, SegFormer, DeepLabV3, EoMT), vision-language modeling (CLIP, SigLIP, SigLIP2), and more. It includes hybrid architectures like MaxViT alongside traditional CNNs and pure transformers. kmodels includes custom layers and backbone support, providing flexibility and efficiency across various applications. For backbones, there are various weight variants like in1k, in21k, fb_dist_in1k, ms_in22k, fb_in22k_ft_in1k, ns_jft_in1k, aa_in1k, cvnets_in1k, augreg_in21k_ft_in1k, augreg_in21k, and many more.

⚡ Installation

From PyPI (recommended)

pip install -U kmodels

From Source

pip install -U git+https://github.com/IMvision12/keras-models

📑 Documentation

Topic	Description
Backbone Models	Classification backbones (ViT, ResNet, Swin, ConvNeXt, EfficientNet, and more) with usage examples and model listing

Segmentation

Model	Description
SAM	Segment Anything Model — promptable segmentation with points, boxes, or masks (ViT-B/L/H)
SAM2	Segment Anything Model 2 — next generation of promptable visual segmentation (Hiera Tiny/Small/Base+/Large)
SAM3	Segment Anything Model 3 — open-vocabulary detection + segmentation with CLIP text encoder (ViT-L/14). Weights require Meta SAM License acceptance on HuggingFace
SegFormer	Transformer-based semantic segmentation with MLP decoder, Cityscapes & ADE20K weights
DeepLabV3	Atrous convolution-based semantic segmentation
EoMT	Encoder-only Mask Transformer for panoptic segmentation

Object Detection

Model	Description
DETR	End-to-end object detection with Transformers (ResNet-50/101 backbones)
RT-DETR	Real-time DETR with ResNet-vd backbone and hybrid encoder (ResNet-18/34/50/101 variants)
RT-DETRv2	RT-DETR v2 with selective multi-scale deformable attention and learnable per-level sampling scale (ResNet-18/34/50/101 variants)
RF-DETR	Real-time detection transformer (Nano, Small, Medium, Base, Large variants)
D-FINE	Fine-grained distribution refinement detector with HGNetV2 backbone (Nano/Small/Medium/Large/XLarge)

Feature Extraction

Model	Description
DINO	Self-supervised ViT-S/B and ResNet-50 backbones trained with self-distillation
DINOv2	Improved self-supervised ViT-S/B/L backbones with LayerScale, trained on LVD-142M

Vision-Language Models

Model	Description
CLIP	Contrastive Language-Image Pre-training for zero-shot classification
SigLIP	Sigmoid loss-based language-image pre-training with multilingual support
SigLIP2	Next-gen SigLIP with improved semantic understanding and 256K vocabulary

📑 Models

• Backbones

  🏷️ Model Name	📜 Reference Paper	📦 Source of Weights
  CaiT	Going deeper with Image Transformers	timm
  ConvMixer	Patches Are All You Need?	timm
  ConvNeXt	A ConvNet for the 2020s	timm
  ConvNeXt V2	ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders	timm
  DeiT	Training data-efficient image transformers & distillation through attention	timm
  DenseNet	Densely Connected Convolutional Networks	timm
  EfficientNet	EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks	timm
  EfficientNet-Lite	EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks	timm
  EfficientNetV2	EfficientNetV2: Smaller Models and Faster Training	timm
  FlexiViT	FlexiViT: One Model for All Patch Sizes	timm
  InceptionNeXt	InceptionNeXt: When Inception Meets ConvNeXt	timm
  Inception-ResNet-v2	Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning	timm
  Inception-v3	Rethinking the Inception Architecture for Computer Vision	timm
  Inception-v4	Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning	timm
  MaxViT	MaxViT: Multi-Axis Vision Transformer	timm
  MiT	SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers	transformers
  MLP-Mixer	MLP-Mixer: An all-MLP Architecture for Vision	timm
  MobileNetV2	MobileNetV2: Inverted Residuals and Linear Bottlenecks	timm
  MobileNetV3	Searching for MobileNetV3	keras
  MobileViT	MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer	timm
  MobileViTV2	Separable Self-attention for Mobile Vision Transformers	timm
  NextViT	Next-ViT: Next Generation Vision Transformer for Efficient Deployment in Realistic Industrial Scenarios	timm
  PiT	Rethinking Spatial Dimensions of Vision Transformers	timm
  PoolFormer	MetaFormer is Actually What You Need for Vision	timm
  Res2Net	Res2Net: A New Multi-scale Backbone Architecture	timm
  ResMLP	ResMLP: Feedforward networks for image classification with data-efficient training	timm
  ResNet	Deep Residual Learning for Image Recognition	timm
  ResNetV2	Identity Mappings in Deep Residual Networks	timm
  ResNeXt	Aggregated Residual Transformations for Deep Neural Networks	timm
  SENet	Squeeze-and-Excitation Networks	timm
  Swin Transformer	Swin Transformer: Hierarchical Vision Transformer using Shifted Windows	timm
  Swin Transformer V2	Swin Transformer V2: Scaling Up Capacity and Resolution	timm
  VGG	Very Deep Convolutional Networks for Large-Scale Image Recognition	timm
  ViT	An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale	timm
  Xception	Xception: Deep Learning with Depthwise Separable Convolutions	keras

• Object Detection

  🏷️ Model Name	📜 Reference Paper	📦 Source of Weights
  D-FINE	D-FINE: Redefine Regression Task of DETRs as Fine-grained Distribution Refinement	transformers
  DETR	End-to-End Object Detection with Transformers	transformers
  RT-DETR	DETRs Beat YOLOs on Real-time Object Detection	transformers
  RT-DETRv2	RT-DETRv2: Improved Baseline with Bag-of-Freebies for Real-Time Detection Transformers	transformers
  RF-DETR	RF-DETR: Neural Architecture Search for Real-Time Detection Transformers	rfdetr

• Segmentation

  🏷️ Model Name	📜 Reference Paper	📦 Source of Weights
  DeepLabV3	Rethinking Atrous Convolution for Semantic Image Segmentation	torchvision
  EoMT	Your ViT is Secretly an Image Segmentation Model	transformers
  SAM	Segment Anything	transformers
  SAM2	SAM 2: Segment Anything in Images and Videos	transformers
  SAM3	SAM 3	transformers (gated)
  SegFormer	SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers	transformers

• Feature Extraction

  🏷️ Model Name	📜 Reference Paper	📦 Source of Weights
  DINO	Emerging Properties in Self-Supervised Vision Transformers	torch.hub
  DINOv2	DINOv2: Learning Robust Visual Features without Supervision	transformers

• Vision-Language-Models (VLMs)

  🏷️ Model Name	📜 Reference Paper	📦 Source of Weights
  CLIP	Learning Transferable Visual Models From Natural Language Supervision	transformers
  SigLIP	Sigmoid Loss for Language Image Pre-Training	transformers
  SigLIP2	SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Features	transformers

📜 License

This project leverages timm and transformers for converting pretrained weights from PyTorch to Keras. For licensing details, please refer to the respective repositories.

• 🔖 kmodels Code: This repository is licensed under the Apache 2.0 License.

🌟 Credits

• The Keras team for their powerful and user-friendly deep learning framework
• The Transformers library for its robust tools for loading and adapting pretrained models
• The pytorch-image-models (timm) project for pioneering many computer vision model implementations
• All contributors to the original papers and architectures implemented in this library

Citing

BibTeX

@misc{gc2025kmodels,
  author = {Gitesh Chawda},
  title = {Keras Models},
  year = {2025},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/IMvision12/keras-models}}