It is an implemetation of attention-based sampler in TASN.
It's based on MobulaOP, and you don't need to re-build MXNet.
In addition, the implementation of attention-based sampler is available for MXNet and PyTorch.
- Install MobulaOP
# Clone the project
git clone https://github.com/wkcn/MobulaOP
# Enter the directory
cd MobulaOP
# Install MobulaOP
pip install -v -e .- Clone TASN project
git clone https://github.com/researchmm/tasn
cd tasn/tasn-mxnet/example/tasn- Clone this project
git clone https://github.com/wkcn/AttentionSamplerThe directory shows as follow:
├─AttentionSampler
│ ├── attention_sampler
│ ├── imgs
│ └── test.py
├── common
├── data
├── init.sh
├── install.sh
├── model
├── model.py
├── readme
├── train.py
└── train.sh- Copy the following code on the head of
model.pyof TASN
import mxnet as mx
import mobula
from AttentionSampler.attention_sampler import attsampler_mx
mobula.op.load('./AttentionSampler/attention_sampler')You can train TASN model now. Enjoy it!
If this project is helpful, Hope to follow me and star the MobulaOP project.
Thank you!
@inproceedings{zheng2019looking,
title={Looking for the Devil in the Details: Learning Trilinear Attention Sampling Network for Fine-grained Image Recognition},
author={Zheng, Heliang and Fu, Jianlong and Zha, Zheng-Jun and Luo, Jiebo},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={5012--5021},
year={2019}
}