2022-paper-fun / lca-cnn

LCA-CNN

Code for paper "Learning Cascade Attention for Fine-grained Image Classification", which currently under review at Elsevier Journal of Neural Networks(NN).

File description

1. bootstrap.py
   Train/test split for CUB-200-2011 dataset.

2. image_rotate.py
   Rotate training set for data augmentation.

3. model.py
   Core model code.

4. inception_train.py
   Training and validation code.

5. inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5
   The iNaturalist pre-train parameters (converted from Link).

6. weights1/weights-gatp-two-stream-inception_v3-006-0.9080.hdf5
   Trained model for CUB-200-2011 dataset. (for reproduction).
   To unzip file:

zip -s- weights1.zip -O weights111.zip
unzip weights111.zip

Dependencies:

• Python (3.6)
• Keras (2.1.5)
• Tensorflow (1.10.0)

---