Image Classification

I built a convolutional neural network that can predict whether two shoes are from the same pair or from two different pairs.

The Data

• There are three main folders: train, test_w and test_m.
• The dataset is comprised of triplets of pairs, where each such triplet of image pairs was taken in a similar setting (by the same person).
• Every image is a numpy array of shape [224, 224, 4]

Data Preparation

• Reshape the data in the shape we want: [N, 3, 2, 224, 224, 3]
  • N - the number of triplets allocated to train, valid, or test
  • 3 - the 3 pairs of shoe images in that triplet
  • 2 - the left/right shoes
  • 224 - the height of each image
  • 224 - the width of each image
  • 3 - the colour channels
• Labelled training data, with same and different pair (positive and negative examples).
• Creating the correct numpy by concatenated the right and left shoe , and for negative 2 difeerent one

CNN Model

• CNN model in PyTorch called CNN that will take images of size 3×448×224 , and classify whether the images contain shoes from the same pair or from different pairs.
• We can find the model in the Function.py file:
  • Conve2d
  • padding and stride
  • Linear
  • Dropout
  • BatchNorm2d
  • relu
  • max_pool2d

Channeled CCN

• I will first manipulate the image so that the left and right shoes images are concatenated along the channel dimension.
• The input to the first convolutional layer will have 6 channels instead of 3 (input shape 6×224×224 ).

Traning

• Criteria : Cross entorpy loss
• Optimizer : Adam
• Hytper paramter : Learning rate , step size , gamma , weight decay -> has been chosen by simulation result
• Batch size of 32 and epochs of 30
• For each epoch and batch size we train the model, with backward and step function, and our goal is to minimize the loss function
• Validate our traning each epoch to estimate the correctness of the traning curve