Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/stevenrice99/facial-expression-recognition

Kaggle Challenges in Representation Learning: Facial Expression Recognition Challenge
https://github.com/stevenrice99/facial-expression-recognition

ai artificial-intelligence classification convolutional-neural-networks deep-learning deep-neural-networks face-recognition facial-recognition kaggle machine-learning ml neural-network neural-networks python python3 pytorch resnet resnet-18 torch torchvision

Last synced: 8 days ago
JSON representation

Kaggle Challenges in Representation Learning: Facial Expression Recognition Challenge

Awesome Lists containing this project

README 

        
# Facial Expression Recognition



This is a solution for the [Kaggle Challenges in Representation Learning: Facial Expression Recognition Challenge](https://www.kaggle.com/competitions/challenges-in-representation-learning-facial-expression-recognition-challenge "Kaggle Challenges in Representation Learning: Facial Expression Recognition Challenge") comparing normal, fully augmented, and a gradual addition of data augmentation during the training process across a simple CNN and a ResNet18 based model.



| ![Normal Data](Data-Normal.png "Normal Data") | 

|:--:| 

| *Normal Data Sample* |



| ![Augmented Data](Data-Augmented.png "Augmented Data") | 

|:--:| 

| *Fully Augmented Data Sample* |



- [Setup](#setup "Setup")

- [Usage](#usage "Usage")

- [References](#references "References")



# Setup



1. Install [Python](https://www.python.org "Python").

   1. Python 3.10.7 was used for this project's creation, but any newer version should work.

2. Install [Git Large File Storage](https://git-lfs.com "Git Large File Storage").

3. Clone this repository.

4. Download the dataset.

   1. Go [here on the Kaggle challenge](https://www.kaggle.com/competitions/challenges-in-representation-learning-facial-expression-recognition-challenge/data?select=fer2013.tar.gz "Kaggle Dataset").

   2. Download "fer2013.tar.gz".

   3. Extract "fer2013.tar.gz". You will likely need to do this twice unless the software you are using automatically recognizes it needs to be done twice, first to get to "fer2013.tar" and then extract it to get the folder "fer2013".

   4. Inside the folder "fer2013", copy the file "fer2013.csv" into the root of your project directory.

   5. Rename "fer2013.csv" to "Data.csv".

5. Install required Python packages. If unfamiliar with Python, "pip" comes with standard Python installations, and you can run "pip *package*" to install a package.

   1. [PyTorch](https://pytorch.org "PyTorch") and TorchVision.

      1. It is recommended you visit [PyTorch's get started page](https://pytorch.org/get-started/locally "PyTorch Get Started") which will allow you to select to install CUDA support if you have an Nvidia GPU. This will give you a command you can copy and run to install PyTorch, TorchVision, and TorchAudio, but feel free to remove TorchAudio from the command as it is not needed. Check which version of CUDA your Nvidia GPU supports. You can check your CUDA version by running "nvidia-smi" from the command line.

      2. When running the script, a message will be output to the console if it is using CUDA.

   2. The remaining packages can be installed via "pip *package*" in the console:

      1. [numpy](https://numpy.org "numpy")

      2. [pandas](https://pandas.pydata.org "pandas")

      3. [tqdm](https://github.com/tqdm/tqdm "tqdm")



# Usage



1. Run "main.py" with the following optional parameters:

   1. -e, --epoch - The number of epochs to train for. Defaults to 100.

   2. -b, --batch - Training and testing batch size. Defaults to 64.

2. Once done, inside the "Models" folder, files with their given name and training modes can be found as follows:

   1. PT file which contains the best weights and bias saved which can be loaded for inference as well as to continue training later.

   2. TXT file which contains an overview of the model. 

   3. CSV file which contains the loss and accuracy for each training epoch.



# References



- "Challenges in Representation Learning: A report on three machine learning contests." I Goodfellow, D Erhan, PL Carrier, A Courville, M Mirza, B Hamner, W Cukierski, Y Tang, DH Lee, Y Zhou, C Ramaiah, F Feng, R Li, X Wang, D Athanasakis, J Shawe-Taylor, M Milakov, J Park, R Ionescu, M Popescu, C Grozea, J Bergstra, J Xie, L Romaszko, B Xu, Z Chuang, and Y. Bengio. arXiv 2013.

- Helper CSV parsing methods based upon [Kaggle submission by Rico Hoffman](https://www.kaggle.com/code/drcapa/facial-expression-eda-cnn "Rico Hoffman Challenges in Representation Learning: Facial Expression Recognition Challenge").