SPA-Net

PyTorch implementation for SPA-Net.

⚙️ Quick Start: Train & Test

🔥 Download fine-tuned model for training Inception-V3 on ILSVRC.

• GoogleDrive
• After downloading, ensure inceptionv3_cls.zip in snapshots/ilsvrc/ and execute following:

unzip inceptionv3_cls.zip

• The model file will be unzipped in this directory snapshots/ilsvrc/inceptionv3_cls/

🔥 Download fine-tuned model for training VGG16 on ILSVRC.

• BaiduNetdisk code：6ih2
• GoogleDrive
• After downloading, put the model into snapshots/ilsvrc/vgg16_spa_#1/

Switch to the scripts directory and execute the following shell code:

🔥 Train

bash ./train_vgg_cub.sh  # For VGG-16 trained on CUB-200-2011
bash ./train_inceptionv3_cub.sh  # For Inception V3 trained on CUB-200-2011

bash ./train_vgg_ilsvrc.sh  # For VGG-16 trained on ILSVRC
bash ./train_inceptionv3_ilsvrc.sh  # For Inception V3 trained on ILSVRC

🔥 Test

bash ./test_vgg_cub.sh  # For VGG-16 test on CUB-200-2011
bash ./test_inceptionv3_cub.sh  # For Inception V3 test on CUB-200-2011

bash ./test_vgg_ilsvrc.sh  # For VGG-16 test on ILSVRC
bash ./test_inceptionv3_ilsvrc.sh  # For Inception V3 test on ILSVRC

🔍 For a more detailed explanation of each parameter, please refer to the HELP.md.