In this project, we will use the CelebA dataset (http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html), which is available on Kaggle. You can download dataset from Kaggle (https://www.kaggle.com/jessicali9530/celeba-dataset):
The dataset contains:
202,599 number of face images of various celebrities
10,177 unique identities, but names of identities are not given
40 binary attribute annotations per image
5 landmark locations
imgalignceleba.zip: All the face images, cropped and aligned
listevalpartition.csv: Recommended partitioning of images into training, validation, testing sets. Images 1-162770 are training, 162771-182637 are validation, 182638-202599 are testing
listbboxceleba.csv: Bounding box information for each image. "x1" and "y1" represent the upper left point coordinate of bounding box. "width" and "height" represent the width and height of bounding box
listlandmarksalign_celeba.csv: Image landmarks and their respective coordinates. There are 5 landmarks: left eye, right eye, nose, left mouth, right mouth
listattrceleba.csv: Attribute labels for each image. There are 40 attributes. "1" represents positive while "-1" represents negative
Using Inception V3 model I am using a pretrained Inception V3 model for which I will retrain some layers and fix the first layers. I will also attach new output layers to perform the new classification task.
As my target variable, I only use the gender feature available in the dataset and detect if the image shows a man or a woman.
Importing libraries and reading Data
Data Visualization
Separating Training, Testing and Validation Data
Data Augmentation
Importing Inception V3 model
Adding Custom Layers
Create and compile final model
Plotting loss functiona and accuracy through epochs
Prediction on Testing Data to check accuracy
Generating new predictions
- The inception model (https://keras.io/api/applications/inceptionv3/) is available from Keras. We can either download pre-trained model or else can connect the notebook with the dataset using "imagenet" as weights.
- Keras ImageDataGenerator and flow_from_dataframe is used to avoid loading all images in memory and fit the model on the entire dataset.