Abstract:
This thesis probes how Machine Learning (ML) can be used to distinguish between original
and counterfeit brake shoe. The research uses a dataset of images collected from different cities
(Islamabad, Lahore, Faisalabad, Rawalpindi and Multan). There are two types of brake shoe
images in the dataset: original and counterfeit; the data is labelled into these two classes
accordingly and fed to the ML models used. Three specific machine learning models,
Constitutional Neural Network (CNN), Multi-layer Perceptron (MLP) and VGG 19 are applied
for distinguishing genuine and fake products. By using a dataset that represents a vast variety
of brake shoes in local markets, the models are made more robust and adaptable. Moreover,
the research methodology involves an examination of the architecture, training process and
evaluation metrics of each model. Confusion matrices are used to evaluate how well the models
can differentiate between fake and original brake shoe. Additionally, accuracy measures such
as precision, recall and F1 score are calculated to offer an assessment of model performance.
Furthermore, a comparative analysis is carried out to determine which model is most effective
at distinguishing two classes of brake shoes. The findings show that the VGG 19 model
surpasses the CNN and MLP models in terms of accuracy and reliability in classification tasks.
VGG19 showed an accuracy of 0.92 compared to 0.90 and 0.81 of CNN and MLP respectively.
For precision as well VGG19 outnumbered other models with score 0f 0.93 in comparison to
0.84 and 0.90 of MLP and CNN respectively. Similarly recall and F-1 score of VGG19 was
higher than other two models used. These results affirmed the usage of VGG19 for usage in
future works. These approaches help in distinguishing between real and fake brake shoes,
which in turn is very effective towards anti-counterfeiting. In general, this study sheds light on
how machine learning can be applied to combat counterfeit products specifically focusing on
brake shoes. The results offer insights for industry players, policymakers and researchers
working on counterfeiting initiatives highlighting the significance of utilizing advanced
technologies to tackle this global issue.
