Choose if you need Keras-compatible pretrained classification models; different framework, so not drop-in for PyTorch.
Classification models trained on ImageNet. Keras.
- 1.2k
- Python
- MIT
pretrained image classification models
A PyTorch collection of pretrained ConvNet models (NASNet, ResNeXt, ResNet, Inception, Xception, DPN, etc.) with a unified API for transfer learning.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose if you need Keras-compatible pretrained classification models; different framework, so not drop-in for PyTorch.
Classification models trained on ImageNet. Keras.
Choose if you need pretrained models in TensorFlow; different framework and API.
High level network definitions with pre-trained weights in TensorFlow
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Stale PyTorch zoo; choose if you also need video/GAN models, but expect no maintenance.
Pretrained Image & Video ConvNets and GANs for PyTorch: NASNet, ResNeXt (2D + 3D), ResNet (2D + 3D), InceptionV4, InceptionResnetV2, Xception, DPN, NonLocalNets, R(2+1)D nets, MultiView CNNs, Temporal Relation Networks, BigGANs StyleGANs, etc.
Narrower DPN-only pretrained weights; choose if you only need DPN and want the original MXNet port, but it's stale.
Dual Path Networks (DPN) supporting pretrained weights converted from original MXNet implementation
Single NASNet model in TensorFlow; choose only if you need NASNet in TF and don't mind old code.
A nasnet in tensorflow
Keras NASNet with weights; narrower alternative if you specifically need NASNet in Keras.
"NASNet" models in Keras 2.0+ with weights
Keras InceptionV4 with pretrained weights; narrow alternative for that one model.
Keras Implementation of Google's Inception-V4 Architecture (includes Keras compatible pre-trained weights)
These projects were analysed and named pretrained-models.pytorch among their alternatives. The relationship is not symmetric — how pretrained-models.pytorch rates them is a separate judgement, made when pretrained-models.pytorch is analysed in its own right.
calls pretrained-models.pytorch “Older pretrained ConvNet zoo; choose if you need its original weight APIs, but it is unmaintained.”
The largest collection of PyTorch image encoders / backbones. Including train, eval, inference, export scripts, and pretrained weights -- ResNet, ResNeXT, EfficientNet, NFNet, Vision Transformer (ViT), MobileNetV4, MobileNet-V3 & V2, RegNet, DPN, CSPNet, Swin Transformer, MaxViT, CoAtNet, ConvNeXt, and more