{"id":11582,"date":"2024-09-30T15:07:43","date_gmt":"2024-09-30T19:07:43","guid":{"rendered":"https:\/\/blogs.mathworks.com\/student-lounge\/?p=11582"},"modified":"2024-10-17T09:18:25","modified_gmt":"2024-10-17T13:18:25","slug":"join_ieee_sp_cup_2025","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/student-lounge\/2024\/09\/30\/join_ieee_sp_cup_2025\/","title":{"rendered":"Join the IEEE SP Cup 2025: Deepfake Face Detection In The Wild Challenge!"},"content":{"rendered":"
\n
<\/div>\n
<\/div>\n
For today’s blog post, Liping Wang joins us to talk about the IEEE SP Cup 2025 and how to kick off your project on deepfake face detection in MATLAB with the starter code. Over to you, Liping…<\/span><\/div>\n
<\/div>\n
<\/div>\n
Hello, future innovators and AI enthusiasts! Are you ready to dive into the world of deepfakes and make your mark in the field of signal processing? We at MathWorks<\/a> invite you to participate in the IEEE Signal Processing Cup<\/a> challenge in 2025, “Deepfake Face Detection In The Wild<\/a>” (DFWild-Cup). This provides you a great chance to tackle real-world problems using cutting-edge AI techniques.<\/div>\n
<\/div>\n
\"\"<\/div>\n
<\/div>\n
Why Is This Challenge Important?<\/span><\/div>\n
With the rise of synthetic data generation, deepfakes have become a significant threat, capable of manipulating public opinion and even leading to identity theft. This challenge is your opportunity to develop methods to identify whether facial images are real or fake, using data captured in diverse, real-world scenarios.<\/div>\n
What\u2019s in It for You?<\/span><\/div>\n
Participating in this challenge not only allows you to apply your skills to a pressing global issue but also gives you a chance to compete for a US$5,000<\/span> grand prize at the IEEE ICASSP 2025<\/a>, the world\u2019s largest technical conference on signal processing. Imagine presenting your work on such a prestigious platform!<\/div>\n
Ready to Get Started?<\/span><\/div>\n
We’ve prepared a MATLAB starter code to help you kick off your project. Here\u2019s a quick guide on how to set up your environment and start experimenting with deepfake detection! To request your complimentary MATLAB license and access additional learning resources, please visit our website<\/a>. You also can find our self-paced online courses on MATLAB and AI at MATLAB Academy<\/a>.<\/div>\n
\n
Table of Contents<\/span><\/div>\n<\/div>\n
Load Data
\n<\/a>Create Image Datastores
\n<\/a>Load or Create a Network
\n<\/a>Prepare Data for Training
\n<\/a>Train Neural Network
\n<\/a>Test Neural Network
\n<\/a>Create submissions
\n<\/a>Conclusion<\/a><\/div>\n
<\/div>\n

Load Data<\/h2>\n
First things first, register your team<\/a> and then find the instructions on how to download the training and validation datasets. Store the archives in a subfolder named datasetArchives<\/span> in your current directory.<\/div>\n
\"\"<\/div>\n
The code below will help you unzip the archives and organize the datasets into ‘real’ and ‘fake’ categories:<\/div>\n
\n
\n
datasetArchives = fullfile(pwd,“datasetArchives”<\/span>);<\/span><\/div>\n<\/div>\n
\n
datasetsFolder = fullfile(pwd,“datasets”<\/span>);<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
if <\/span>~exist(datasetsFolder,‘dir’<\/span>)<\/span><\/div>\n<\/div>\n
\n
mkdir(datasetsFolder);<\/span><\/div>\n<\/div>\n
\n
untar(fullfile(datasetArchives,“train_fake.tar”<\/span>),fullfile(datasetsFolder,“train”<\/span>));<\/span><\/div>\n<\/div>\n
\n
untar(fullfile(datasetArchives,“train_real.tar”<\/span>),fullfile(datasetsFolder,“train”<\/span>));<\/span><\/div>\n<\/div>\n
\n
untar(fullfile(datasetArchives,“valid_fake.tar”<\/span>),fullfile(datasetsFolder,“valid”<\/span>));<\/span><\/div>\n<\/div>\n
\n
untar(fullfile(datasetArchives,“valid_real.tar”<\/span>),fullfile(datasetsFolder,“valid”<\/span>));<\/span><\/div>\n<\/div>\n
\n
end<\/span><\/span><\/div>\n<\/div>\n<\/div>\n

Create Image Datastores<\/h2>\n
Image datastores are essential for handling large collections of images efficiently. An image datastore allows you to store extensive collections of image data, including those that exceed memory capacity, and efficiently read image batches during neural network training.<\/div>\n
Here\u2019s how you can set them up for your training and validation datasets. You need to specify the folder with the extracted images and indicate that the subfolder names correspond to the image labels in the function imageDatastore<\/span>, and then shuffle<\/span> the images.<\/div>\n
\n
\n
trainImdsFolder = fullfile(datasetsFolder,‘train’<\/span>);<\/span><\/div>\n<\/div>\n
\n
validImdsFolder = fullfile(datasetsFolder,‘valid’<\/span>);<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
imdsTrain = shuffle(imageDatastore(trainImdsFolder, …<\/span><\/span><\/div>\n<\/div>\n
\n
IncludeSubfolders=true, …<\/span><\/span><\/div>\n<\/div>\n
\n
LabelSource=“foldernames”<\/span>));<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
imdsValid = shuffle(imageDatastore(validImdsFolder, …<\/span><\/span><\/div>\n<\/div>\n
\n
IncludeSubfolders=true, …<\/span><\/span><\/div>\n<\/div>\n
\n
LabelSource=“foldernames”<\/span>));<\/span><\/div>\n<\/div>\n<\/div>\n
By checking the size of files in the image data stores, you can see the training data store contains 262160 images while the validation one contains 3072 images. Since we do not have a test dataset for evaluating the performance now, we use the splitEachLabel<\/span> function to partition the training image datastore into two new datastores, i.e. 10% for training and 2% for testing.<\/div>\n
\n
\n
[imdsTrain,imdsTest] = splitEachLabel(imdsTrain,0.1,0.02,“randomized”<\/span>);<\/span><\/div>\n<\/div>\n<\/div>\n
Now let us get the class names and the number of classes, and then display some sample facial images as follows.<\/div>\n
\n
\n
classNames = categories(imdsTrain.Labels);<\/span><\/div>\n<\/div>\n
\n
numClasses = numel(classNames);<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
numImages = numel(imdsTrain.Labels);<\/span><\/div>\n<\/div>\n
\n
idx = randperm(numImages,16);<\/span><\/div>\n<\/div>\n
\n
I = imtile(imdsTrain,Frames=idx);<\/span><\/div>\n<\/div>\n
\n
figure<\/span><\/div>\n<\/div>\n
\n
imshow(I)<\/span><\/div>\n<\/div>\n<\/div>\n
\"\"<\/div>\n

Load or Create a Network<\/h2>\n
Now that your data is ready, the next step is to load a pre-trained network or create a new one. Using a pre-trained network like ResNet or VGG can save time and improve performance, especially if you’re new to deep learning. MATLAB provides several pre-trained models you can use as a starting point.<\/div>\n
Here’s a simple way to load a pre-trained network. As an example, we use the function imagePretrainedNetwork<\/span> to load a pre-trained ResNet-50 neural network with a specified number of classes. Note that you need to install the addon “Deep Learning Toolbox Model for ResNet-50 Network<\/a>” in advance of running the code.<\/div>\n
\n
\n
net = imagePretrainedNetwork(“resnet50”<\/span>,NumClasses=numClasses);<\/span><\/div>\n<\/div>\n<\/div>\n
If you prefer to create your own network, MATLAB\u2019s Deep Network Designer<\/a> app is a great tool to design and visualize your model. You can also Import PyTorch\u00ae Model Using Deep Network Designer<\/a>.<\/div>\n

Prepare Data for Training<\/h2>\n
Preparing your data involves resizing images to match the input size of your network and augmenting them to improve model robustness. MATLAB makes it easy with built-in functions including imageDataAugmenter<\/span> and augmentedImageDatastore<\/span>.<\/div>\n
Data augmentation techniques like rotation, scaling, and flipping can help make your model more generalizable. Here we perform additional augmentation operations including randomly flipping the training images along the vertical axis and randomly translating them up to 30 pixels horizontally and vertically on the training images.<\/div>\n
\n
\n
inputSize = net.Layers(1).InputSize;<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
pixelRange = [-30 30];<\/span><\/div>\n<\/div>\n
\n
imageAugmenter = imageDataAugmenter( …<\/span><\/span><\/div>\n<\/div>\n
\n
RandYReflection=true, …<\/span><\/span><\/div>\n<\/div>\n
\n
RandXTranslation=pixelRange, …<\/span><\/span><\/div>\n<\/div>\n
\n
RandYTranslation=pixelRange);<\/span><\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n
\n
augimdsTrain = augmentedImageDatastore(inputSize(1:2),imdsTrain, …<\/span><\/span><\/div>\n<\/div>\n
\n
DataAugmentation=imageAugmenter);<\/span><\/div>\n<\/div>\n<\/div>\n
The validation and the testing images only need to be resized, so you can use an augmented image datastore without specifying any additional preprocessing operations to do the resizing automatically.<\/div>\n
\n
\n
augimdsValid = augmentedImageDatastore(inputSize(1:2),imdsValid);<\/span><\/div>\n<\/div>\n
\n
augimdsTest = augmentedImageDatastore(inputSize(1:2),imdsTest);<\/span><\/div>\n<\/div>\n<\/div>\n

Train Neural Network<\/h2>\n
With your data and network ready, it’s time to train your model.<\/div>\n
To do transfer learning, the last layer with learnable parameters requires retraining. This is usually a fully connected layer or a convolutional layer with an output size that matches the number of classes. To increase the level of updates to this layer and speed up convergence, you can increase the learning rate factor of its learnable parameters using the setLearnRateFactor<\/span> function, i.e. set the learning rate factors of the learnable parameters to 10.<\/div>\n
\n
\n
net = setLearnRateFactor(net,“res5c_branch2c\/Weights”<\/span>,10);<\/span><\/div>\n<\/div>\n
\n
net = setLearnRateFactor(net,“res5c_branch2c\/Bias”<\/span>,10);<\/span><\/div>\n<\/div>\n<\/div>\n
Next, define training options such as the optimizer, learning rate, and number of epochs. Your choices require empirical analysis. You can use the Experiment Manager<\/a> app to explore different training options with experiments. As an example, we set the training options as follows:<\/div>\n