Project Information

Objective

To explore the application of data manipulation, transfer learning and built a custom deep neural network for a cutlery image dataset and then perform model evaluation to select the hero model to classify cutlery images into four categories - cup, knife, fork, spoon.

  • Techniques:
  1. Keras image preprocessing
  2. Data augmentation
  3. Dataset prefetching
  4. Reusing weights of neural network
  5. Hyperparameter tuning
  6. Model optimization and evaluation