Preamble: This notebook was adapted from an Introduction to Machine Learning course assignment. This notebook is intended to demonstrate a basic understanding of convolutional neural network architecture.
Goal: To develop an emotion detection CNN-based model that can detect emotions from images in real-time.
Instructions: The entire full notebook should run less than 35 minutes to run on T4 GPU. The solutions for this notebook are self-contained. This notebook was intended to be executed and tested on Google Colab. Using the commands:
runtime
→run all
.
Rationale: There are some general understanding questions and interpretations of the model placed in markdown cells with this emoji (❓) and answers placed in cells with this emoji (✅). This will provide some context on the application and an explanation of some of the design decisions made.
Dataset: Facial Expression Recognition 2013 Dataset. The required dataset is located within the
FER-2013.rar
file. Download it and store it inGoogle Drive
. It should be underMy Drive
. Note: After uploading it, check the location by right-clicking on the fileFile Information
→Details
→Location
.