/Computational_Intelligence_AUT

Projects done in this course

Primary LanguageJupyter Notebook

Computational_Intelligence_AUT

HW1

  • implementing a simple single neural network using sigmoid and tanh and comparing results

HW2 & HW3

  • Implementing a MLP , based on differenet datasets

HW4

  • Using K-means algorithm in blobs dataset and image size reduction

HW5

  • Implementing RBFN

Final Project

As a final project of this course we were asked to built a network which must have ability to detect the face area and then recognising the person. We used 14 Celebs dataset on Kaggle. We used densenet and then Resnet as our base model and changed the last layer in different ways to fine-tune the model for our specific problem. We experiment with different combinations of the last layer and choose the best one based on the validation accuracy. And finally to get the smaller and more efficient network we used MobileNet as our based model. Here are summaries of diffrent model we used.

  • ResNet
Screen Shot 1402-02-21 at 23 53 20
  • DenseNet
Screen Shot 1402-02-21 at 23 55 47
  • MobileNet
Screen Shot 1402-02-21 at 23 56 35
  • Number of Parameter and Size of Model
Screen Shot 1402-02-21 at 23 57 47

Contributers in project

  • Reyhaneh Ahani
  • Mahdiye Sadat Benis
  • Samira Saljoghi
  • Fatemeh Rafiee