Skip to content

Latest commit

 

History

History
49 lines (36 loc) · 2.28 KB

README.md

File metadata and controls

49 lines (36 loc) · 2.28 KB

Multicore-Parallelization-of-Neural-Networks-for-Face-Recognition

License

Multicore Parallelization of Neural Networks for Face Recognition, a neural network built on top of CMU code and has been updated, enhanced, and parallelized. (2013)

Report: https://1drv.ms/b/s!AjpxhCM4OMFHg_APRliCaNCRj-kvBg

Please cite the paper if you use this code or data, and also mention me in your work.

This repo contains:

  • Samples folder: which has 96 images for many people.

  • all.txt: it has list of all the samples. No actual use for such file.

  • Source Code:

    • backprop.c
    • imagenet.c
    • pgmimage.c
    • main.c
    • backprop.h
    • imagenet.h
    • pgmimage.h

  • Input files:

    • samples.txt: this file will be used in series training, it has the list of training samples.
    • samples1.txt: this file will be used in parallel training for thread 1, it has the list of training samples.
    • samples2.txt: this file will be used in parallel training for thread 2, it has the list of training samples.
    • test.txt: this file will be used in both series and parallel training, it has the list of images that will be used in the validation process

  • Output files:

    • Net_hidden_initial_weights.txt: output of series training

    • Net_hidden_final_weights.txt: output of series training

    • Net_input_initial_weights.txt: output of series training

    • Net_input_final_weights.txt: output of series training

    • Net1_hidden_initial_weights.txt: output of parallel training, thread 1

    • Net1_hidden_final_weights.txt: output of parallel training, thread 1

    • Net1_input_initial_weights.txt: output of parallel training, thread 1

    • Net1_input_final_weights.txt: output of parallel training, thread 1

    • Net2_hidden_initial_weights.txt: output of parallel training, thread 2

    • Net2_hidden_final_weights.txt: output of parallel training, thread 2

    • Net2_input_intial_weights.txt: output of parallel training, thread 2

    • Net2_input_final_weights.txt: output of parallel training, thread 2

    • _merged_hidden_weights.txt: output of parallel training, after merging weights of all threads

    • _merged_input_weights.txt: output of parallel training, after merging weights of all threads