Face Detection in a crowded environment is a challenging task especially for fully-automated computer-vision systems, though not for the same reasons as humans. Firstly, a module of the system needs to scan the input image and extract all the faces with their location (face detection). Secondly, ano
Classroom Crowed Attendance System
Face Detection in a crowded environment is a challenging task especially for fully-automated computer-vision systems, though not for the same reasons as humans. Firstly, a module of the system needs to scan the input image and extract all the faces with their location (face detection). Secondly, another module has to compare each extracted face against the image of the target face and return a degree of match (face identification). If the match is above a certain threshold, the image is classified as containing the target. Both face detection and face identification unavoidably produce false-positive results and cause accuracy reduction. Despite these difficulties, research has shown that deep neural networks do not suffer from some limitations of the human visual system. In principle, this means that computer-vision systems could become more accurate than the average human, especially in tasks where visual information is presented for a brief amount of time. Unlike humans, the performance of computer-vision systems does not degrade over time (e.g., due to fatigue). However, it usually degrades way below that of humans when moving from constrained to realistic environments.
This system is targeting the Pakistani society where Male and female students are in facial growth, men hairs and styles of Niqab respectively. So the system is using a face recognition algorithm using AI steps to recognize the student’s attendance with just one single image from their class with high accuracy.
First of all GUI of the system will be created, where admin and coordinator interact with the system and add records like pictures and other data of students. Records save in MYSQL database. The pictures of students are used as a training dataset for the face recognition system. For detecting a face in the crowd first we prepare custom dataset of Pakistani students then image preprocessing and landmark detection of faces from the dataset is completed by using dlib and HOG (Histogram of Oriented Gradients) algorithm. Then face recognition is implemented by using a face recognition library. In this project teacher take a single picture of class and the algorithm is allowed to detect faces and mark the attendance from the picture.
Automatic Attendance can save much time that is comparatively wasted in the traditional attendance management system. The most important feature of the software is that it can recognize a simple human face even in different situations like the same person with a face mask, weight gained, different situations of facial hair growth in men or women with different styles of hijab, etc. It is maybe the first system that focuses on Pakistani community and culture, it will allow promoting the use of technology in Pakistani society, especially in universities where many female students who don’t feel comfortable to take off their Niqab just for their attendance and or face detection, the system will able to detect them with their different Niqab styles and it can be able to detect male students who have different facial growth.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Oppo A52(Ai camera) | Equipment | 1 | 35000 | 35000 |
| Server and hosting charges | Miscellaneous | 1 | 10000 | 10000 |
| Total in (Rs) | 45000 |
Due to the reduction of non-renewable resources like coal, oil, natural gas etc. the...
Construction management is very helpful for managing all the activities which we do during...
Making of machine, that is capable of measuring the dimensions of the part while manufactu...
In Pakistan there is no proper guideline for students to choose their own interested field...
Summary: This project will be design for LED based street lights with an auto-intensity co...