Multi-Task Emoji Detection Model: Class Prediction and Bounding Box Regression

This project involves the development of a deep learning model to detect emojis in images, predict their

class, and localize them using bounding box regression. The model is trained on 144x144 RGB images with

a dataset of 9 unique emojis. It is a multi-task learning setup, combining classification and localization tasks

using a convolutional neural network (CNN). The project also implements custom metrics like Intersection

over Union (IoU) for evaluating bounding box predictions and features a custom data generator for on-the-fly

example creation.