Applied Machine Learning Systems ELEC0132 Assignment

Maryam Habibollahi (SN: 15000241)

• Detection and removal of noisy images
• Training, validation and testing subsets division
• Train ML models to perform
  • Binary
  1. Emotion recognition (smile/!smile)
  2. Age identification (young/old)
  3. Glasses detection (with/without)
  4. Human detection (real/avatar)
  • Multiclass
  5. Hair colour recognition (ginger, blond, brown, grey, black, bald)

Dataset folder

• Classification models: Scikit-learn, Keras, dlib, and OpenCV.

• Data processing, storage, and representation: NumPy, Pandas, Time, OS, and MatplotLib.

• landmarks.py

Implements funtions on Haar Cascode, HOG, and Deep Learning-based face detectors for obtaining image landmarks from detected faces.

• classification.py

Includes the required functions for SVM and MLP implementation, as well as cross-validation testing.

• utils.py

Provides the utility functions for handling files and data used in landmarks.py and classification.py.

• testing.py

Calls the detection and classification functions and stores the results to csv. This file includes one function for binary tasks 1-4 and two for multiclass task 5. The landmarks previously stored in out/ are used by default to save time. To re-run the face detector, uncomment function update_features()

• models/lenet.py

Implements the LeNet architecture for the multiclass classification task

• code/out/

Includes all the results from the tests (see out/README.md) Face detector features are stored in Face_detection/ for convenience

• code/models/

Contains face detection models and the LeNet architecture setup file