• Dr.Rakesh Kumar Giri

Powerful deep learning machine learning models are difficult to apply to low-segment embedded systems and low-segment mobile devices due to memory and battery limitations on platforms with limited memory and battery. This is a big problem. Deep learning models have yet to be deployed in very low-segment embedded systems such as smartphones (low cost), smart wearables, traffic lights, and embedded systems. To solve this problem, this document presents a new face recognition method using his CNN. With their help, trained CNN-based models can be easily deployed on memory-constrained platforms such as low-segment embedded systems and low-memory mobile devices/smart wearables. Training tricks for embedded systems with limited memory. In our tests, we ran the proposed network architecture over a network on a very memory-constrained system (at least 2MB). The results show that when running on a platform with a very small memory segment compared to standard CNN-based models, we get very good results on a variety of face verification datasets, and on embedded devices with limited memory. Shows that you can get a model size of less than 1 MB .

Face recognition, CNN, memory-constrained devices

"MACHINE LEARNING BASED FACE RECOGNITION", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.5, Issue 2, page no.154-160, February-2018, Available :http://www.jetir.org/papers/JETIR1802326.pdf

"MACHINE LEARNING BASED FACE RECOGNITION", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.5, Issue 2, page no. pp154-160, February-2018, Available at : http://www.jetir.org/papers/JETIR1802326.pdf