PyTorch implementation of Spatial Transformer Networks applied to the CelebA dataset

Author: Esteban Leiva Montenegro

Spatial Transformer Networks (STNs) are a powerful class of neural networks that enable spatial manipulation of input data within the network architecture itself¹. With PyTorch, implementing STNs becomes more accessible, offering flexibility and efficiency. In this context, leveraging the CelebA dataset², renowned for its diverse facial images, provides a rich environment for exploring STNs capabilities in tasks like gender recognition implemented with a ResNet101³.

1. Usage

1.1 Download the dataset

Note that the ~200,000 CelebA face image dataset is relatively large (~1.3 Gb). The download link provided below was provided by the author on the official CelebA website at http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html.

1. Download and unzip the file img_align_celeba.zip, which contains the images in jpeg format.

2. Download the list_attr_celeba.txt file, which contains the class labels

3. Download the list_eval_partition.txt file, which contains training/validation/test partitioning info

4. Move the downloaded files to ./celeba directory

       📁 Root
       ├── 📁 celeba
       │   ├── 📜 img_align_celeba.zip
       │   ├── 📜 list_attr_celeba.txt
       │   └── 📜 list_eval_partition.txt
       ├── 📁 src
       │   └── ...
       └── 📜 README.md
   

1.2 Install dependencies

You can install all the project dependencies running the following command pip install -r requirements.txt

1.3. Run ./src/app.py

Tip

To adjust hyperparameters, you can modify the ./src/config.json file. Note: Ensure that the BATCH_SIZE in the loaded data and the model BATCH_SIZE remain consistent.

1. Run the following command python ./src/app.py

References

Footnotes

1. Jaderberg, M., Simonyan, K., Zisserman, A., & Kavukcuoglu, K. (2016). "Spatial Transformer Networks." arXiv preprint arXiv:1506.02025 [cs.CV], version 3, February 4, 2016. [Online]. Available: https://arxiv.org/abs/1506.02025 ↩

2. Liu, Z., Luo, P., Wang, X., & Tang, X. (2015). "Deep Learning Face Attributes in the Wild." In Proceedings of International Conference on Computer Vision (ICCV), December 2015. ↩

3. Raschka, S. (2019). "CNN-ResNet101-CelebA." [Online]. Available: https://github.com/rasbt/deeplearning-models/blob/master/pytorch_ipynb/cnn/cnn-resnet101-celeba.ipynb ↩