Week 1 — Project LEAFS
Introduction
The name of the project comes from a leaf metaphor, this metaphor is built on a tree and leaf relationship. The tree is a reference to a country and the student is a reference to a leaf. A student is as vital for countries' future as a leaf is to a tree. LEAFS stands for Lecture Efficiency Assessment from Footages of Students. We aim to increase the lecture's efficiency by detecting distracted students.
A Brief Overview of the Problem
Nowadays, the attitudes of the students such as sleeping, using technological devices, and taking notes etc, affect the efficiency of the lectures.
Project Overview
The project is built on up detection of 3 positive and 4 negative attitudes of students.
The negative attitudes of the students such as sleeping, using mobile phones, yawning and eating or drinking during the lecture decrease the efficiency of the lecture. On the contrary, taking notes, listening to the lecture and raising hands are considered positive attitudes which increase the efficiency of the lecture.
In the project, we will try to assess the lecture efficiency by using deep learning models for example YOLOv5, ResNet and machine learning algorithms.
The ideology behind the project is to give information to the lecturer about the efficiency of the lecture by giving weight to the student attitudes which are detected from the footage.
Dataset
In this project, the dataset will be built from different sources. These sources are ready-made datasets on GitHub, images scraped from the internet and photos which we took from our friends.,
The dataset that we will collect and combine will have 3 positive and 4 negative class labels. Each class will contain around 1000 images. All the images will be labelled by hand.
Related Dataset: https://github.com/it-maranatha/classroom_dataset
Related Works
The article¹ covers a detection model for student class behaviours. They combine the text features which are extracted from the images and deep learning models.
The article² covers a detection model that uses an improved YOLOv5 algorithm by combining multi-scale feature fusion and attention mechanism.
The article³ and article⁵ cover, simply, the mobile phone usage detection model that uses deep learning techniques.
The article⁴ covers a model that uses machine-learning techniques to detect mobile phone usage in class.
References
- https://link.springer.com/article/10.1007/s00521-020-05587-y
- https://www.researchgate.net/publication/361841996_Classroom_Behavior_Detection_Based_on_Improved_YOLOv5_Algorithm_Combining_Multi-Scale_Feaure_Fusion_and_Attention_Mechanism
- https://www.researchgate.net/publication/343290099_Detecting_Usage_of_Mobile_Phones_using_Deep_Learning_Technique
- https://www.ijert.org/monitoring-mobile-usage-in-classroom
- https://dl.acm.org/doi/pdf/10.1145/3411170.3411275
You can click on the link to reach the second blog:
https://medium.com/@nsergn/week-2-project-leafs-c86511e30790
