Researching the use of four Machine Learning models and approaches. These are ANN, CNN, ResNet-50 and SVM which are applied to two different datasets.
- Pokemon Image Classification
- Car or Truck Image Classification
This project explores the use of computer vision and how it is used within the modern society. Furthermore, the study researches approaches used today to make use of computer vision which is a subset of machine learning, then applying this to datasets for classification. Vast amounts of people use technologies derived from computer vision in their daily lives but are unaware on the techniques used to reach the end result. For example, facial recognition is something that has become part of many of our lives. We see them at airports, on phones and many other places. Also, what other technologies may arise from computer vision that we can make use of today? I believe there is yet a lot to explore and will use computer vision to research on what architectures and neural networks perform the best given different situations. This is what my project aims to answer by providing an insight on the underlying concepts of certain technologies derived from computer vision. Then I will apply this knowledge towards creating four image classification models using two different datasets and determining which performs the best given the task at hand. I aim to make comparisons between different architectures and neural networks.
- Program can be run as a jupyter notebook e.g. google colab
- Modifications such as changing the folder/file names
- Can also modify what classes to classify for the ANN/SVM
- Datasets are required to be added to the correct path or be accessible
Below are links to download my ResNet-50 models which have been trained using Keras.
Datasets.
Other references can be viewed within my report such as code snippets etc.
NOTE: Some models are resource intensive so take this into consideration when training the models.