GKalliatakis/Keras-VGG16-places365

Provides two pre-trained VGG16 variants converted from the original Caffe models: one optimized for 365-category scene classification, and a hybrid variant trained on both Places365 and ImageNet for generic feature extraction. Weights auto-download on instantiation and support both TensorFlow and Theano backends with automatic dimension ordering adaptation. The hybrid model enables transfer learning by exposing intermediate feature layers without the classification head, making it suitable for downstream visual recognition tasks beyond scene recognition.

203 stars. No commits in the last 6 months.