Smart Identity Management System by Face Detection Using Multitasking Convolution Network

Table 1

Existing techniques and their limitations.

Author	Title	Model	Weaknesses
Faruk et al. [20]	“Image to Bengali Caption Generation Using Deep CNN and Bidirectional Gated Recurrent Unit”	Deep Face	When the architecture of CNN was made hybrid with multifarious detectors, it limited the variant part of an object in remarkable identification systems. However, this methodology can only be used to identify a few categories of pictures.
Alimuin et al. [21]	“Deep hypersphere embedding for real-time face recognition”	Hybrid combination of MTCNN and FaceNet	Targets only an aspect of constraints in facial features and face recognition with low accuracy.
Jin et al. [22]	“Face recognition based on MTCNN and FaceNet”	Framework of deep cascading with Neural Network	Fails to consider surrounding intensity levels and distance between camera and face.
Xiao et al. [23]	“The Improvement of MTCNN ALGORITHM: Face Recognition with Mask”	MTCNN	This model does not consider the environmental light intensity levels, surrounding atmospheric conditions, and the distance between camera and subject.