- Who: The main actors in this story are 1) the images of the people, 2) the business side who created the requirements to identify ethnicity, gender, age, and expressions, 3) the data scientist (our team), and 4) the AI apps and tool used.
- The data scientist understands the domain as described in the story telling and data sharing sections. The data has been explored by the data scientist in the application code.
- What : We applied multi layered and multi-output CNNs for age, gender, and ethnicity predictions. The expressions were determined using OpenCV tool with Haar Cscade for mood detection.
- The story involves image analysis. Based on the background studies in Machine Learning and Image Analysis, we determined that CNN serves best for our main ML models.
- When has to do with the iterations (Calibration 2). How much time did it take for experimentation? How efficient is the modeling/algorithm?: We have performed this study in three iterations. First, we defined the scopes of our project. Then we collected responses in two installments from the instructor (the project owner ). Finally we created and ran the application. The training of the model took most of the times based on the choice of Runtime on google colab.
- Can the data scientist determine the acceptance level of the model (validation with accuracy and runtime performance) considering the targeted users?: Both of our models performed reasonably with the best results of gender determination accuracy ~ 88%, ethnicity accuracy ~ 76% and age mean absolute error ~ 10% from the model 1. We further investigated the age determination standard deviations with different age groups as described in our paper.
- Where has to do with the learning environment. Where did this experiential learning process take place?: CS5542 coursework at the UMKC CSEE department.
- Why : Refer to our project report for the details on the models.