Keras | VGG16 Places365

Introducing Keras Application Zoo: A library for reusable deep learning models in Keras.

You have just found the Keras models of the pre-trained VGG16 CNNs on Places365-Standard (~1.8 million images from 365 scene categories).


CNNs trained on Places365 database (latest subset of Places2 Database) could be directly used for scene recognition, while the deep scene features from the higher level layer of CNN could be used as generic features for visual recognition.

The Keras models has been obtained by directly converting the Caffe models provived by the authors (all the original Caffe-based resources can be found there).


This repository contains code for the following Keras models: - VGG16-places365 - VGG16-hybrid1365


All architectures are compatible with both TensorFlow and Theano, and upon instantiation the models will be built according to the image dimension ordering set in your Keras configuration file at ~/.keras/keras.json. For instance, if you have set image_dim_ordering=tf, then any model loaded from this repository will get built according to the TensorFlow dimension ordering convention, “Width-Height-Depth”.

Pre-trained weights can be automatically loaded upon instantiation (weights='places' argument in model constructor for all image models). Weights are automatically downloaded.