SPP layer could be added in CNN model between convolutional layer and fully-connected lay, so that you can input multi-size images into your CNN model. We use this structure in the paper Pedestrian-Synthesis-GAN: Generating Pedestrian Data in Real Scene and Beyond
The function spatial_pyramid_pool() in file spp_layer.py is independent. It could be added in your own models.
See this:Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
If you find this work useful for your research, please cite:
@article{ouyang2018pedestrian,
title={Pedestrian-Synthesis-GAN: Generating Pedestrian Data in Real Scene and Beyond},
author={Ouyang, Xi and Cheng, Yu and Jiang, Yifan and Li, Chun-Liang and Zhou, Pan},
journal={arXiv preprint arXiv:1804.02047},
year={2018}
}
and
@inproceedings{he2014spatial,
title={Spatial pyramid pooling in deep convolutional networks for visual recognition},
author={He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian},
booktitle={European conference on computer vision},
pages={346--361},
year={2014},
organization={Springer}
}