Practical 6
Tunde Ope-Davies
Create a machine learning model for a classification problem of your choice using Teachable Machine
Project Card
Project: Build, train and deploy an Artificial Intelligence (AI) model to male and female doctoral programme applicants
Value Proposition
The Centre for Digital Humanities, University of Lagos sought a doctoral candidates identification system to effectively manage and maintain our doctoral degree programme admissions process.
Data:
· Input: Photos of female and male doctoral candidates
· Output: Doctoral candidate detection status
Endpoint Deployment Link
https://teachablemachine.withgoogle.com/models/jAI9zHo3b/
Reflection
· In seeking to create this model, we defined the objectives clearly viz, seeking to create a smart gate model for the selection of potential doctoral candidates. We wanted to deploy a machine learning model that will differentiate between male and female candidates with high accuracy, providing a versatile tool for selection, administration and research supervision process.
· We collected images of prospective male and female doctoral candidates from google images platform. We created two classes along that parameters in order to properly identify and differentiate classes/gender differentiation of candidates applying to our graduate programme. The challenges faced include difficulty in obtaining balanced, diverse and sufficient data from the internet. Some of the images did not accurately represent prospective doctoral candidates. Many of them are images of graduate students who have already convoked. To overcome the challenge, we widened the scope of our search for relevant images.
· We uploaded the sample images of both male and female candidates and tested the model. Interestingly, it returned an average of 81% accuracy for males and 97-100% for females.
· The deployment of the model yielded a good result to confirm the proposition that we tested.
· The exercise has enhanced my skills in AI.