/keras-applications

Reference implementations of popular deep learning models.

Primary LanguagePythonOtherNOASSERTION

Keras Applications

Build Status

Keras Applications is the applications module of the Keras deep learning library. It provides model definitions and pre-trained weights for a number of popular archictures, such as VGG16, ResNet50, Xception, MobileNet, and more.

Read the documentation at: https://keras.io/applications/

Keras Applications may be imported directly from an up-to-date installation of Keras:

from keras import applications

Keras Applications is compatible with Python 2.7-3.6 and is distributed under the MIT license.

Performance

  • The top-k accuracies were obtained using Keras Applications with the TensorFlow backend on the 2012 ILSVRC ImageNet validation set and may slightly differ from the original ones.
    • Input: input size fed into models
    • Top-1: single center crop, top-1 accuracy
    • Top-5: single center crop, top-5 accuracy
    • Size: rounded the number of parameters when include_top=True
    • Stem: rounded the number of parameters when include_top=False
Input Top-1 Top-5 Size Stem References
VGG16 224 71.268 90.050 138.4M 14.7M [paper] [tf-models]
VGG19 224 71.256 89.988 143.7M 20.0M [paper] [tf-models]
ResNet50 224 74.928 92.060 25.6M 23.6M [paper] [tf-models] [torch] [caffe]
ResNet101 224 76.420 92.786 44.7M 42.7M [paper] [tf-models] [torch] [caffe]
ResNet152 224 76.604 93.118 60.4M 58.4M [paper] [tf-models] [torch] [caffe]
ResNet50V2 299 75.960 93.034 25.6M 23.6M [paper] [tf-models] [torch]
ResNet101V2 299 77.234 93.816 44.7M 42.6M [paper] [tf-models] [torch]
ResNet152V2 299 78.032 94.162 60.4M 58.3M [paper] [tf-models] [torch]
ResNeXt50 224 77.740 93.810 25.1M 23.0M [paper] [torch]
ResNeXt101 224 78.730 94.294 44.3M 42.3M [paper] [torch]
InceptionV3 299 77.898 93.720 23.9M 21.8M [paper] [tf-models]
InceptionResNetV2 299 80.256 95.252 55.9M 54.3M [paper] [tf-models]
Xception 299 79.006 94.452 22.9M 20.9M [paper]
MobileNet(alpha=0.25) 224 51.582 75.792 0.5M 0.2M [paper] [tf-models]
MobileNet(alpha=0.50) 224 64.292 85.624 1.3M 0.8M [paper] [tf-models]
MobileNet(alpha=0.75) 224 68.412 88.242 2.6M 1.8M [paper] [tf-models]
MobileNet(alpha=1.0) 224 70.424 89.504 4.3M 3.2M [paper] [tf-models]
MobileNetV2(alpha=0.35) 224 60.086 82.432 1.7M 0.4M [paper] [tf-models]
MobileNetV2(alpha=0.50) 224 65.194 86.062 2.0M 0.7M [paper] [tf-models]
MobileNetV2(alpha=0.75) 224 69.532 89.176 2.7M 1.4M [paper] [tf-models]
MobileNetV2(alpha=1.0) 224 71.336 90.142 3.5M 2.3M [paper] [tf-models]
MobileNetV2(alpha=1.3) 224 74.680 92.122 5.4M 3.8M [paper] [tf-models]
MobileNetV2(alpha=1.4) 224 75.230 92.422 6.2M 4.4M [paper] [tf-models]
MobileNetV3(small) 224 68.076 87.800 2.6M 0.9M [paper] [tf-models]
MobileNetV3(large) 224 75.556 92.708 5.5M 3.0M [paper] [tf-models]
DenseNet121 224 74.972 92.258 8.1M 7.0M [paper] [torch]
DenseNet169 224 76.176 93.176 14.3M 12.6M [paper] [torch]
DenseNet201 224 77.320 93.620 20.2M 18.3M [paper] [torch]
NASNetLarge 331 82.498 96.004 93.5M 84.9M [paper] [tf-models]
NASNetMobile 224 74.366 91.854 7.7M 4.3M [paper] [tf-models]
EfficientNet-B0 224 77.190 93.492 5.3M 4.0M [paper] [tf-tpu]
EfficientNet-B1 240 79.134 94.448 7.9M 6.6M [paper] [tf-tpu]
EfficientNet-B2 260 80.180 94.946 9.2M 7.8M [paper] [tf-tpu]
EfficientNet-B3 300 81.578 95.676 12.3M 10.8M [paper] [tf-tpu]
EfficientNet-B4 380 82.960 96.260 19.5M 17.7M [paper] [tf-tpu]
EfficientNet-B5 456 83.702 96.710 30.6M 28.5M [paper] [tf-tpu]
EfficientNet-B6 528 84.082 96.898 43.3M 41.0M [paper] [tf-tpu]
EfficientNet-B7 600 84.430 96.840 66.7M 64.1M [paper] [tf-tpu]

Reference implementations from the community

Object detection and segmentation

Sequence learning

Reinforcement learning

Generative adversarial networks