You are on page 1of 4

Khoule Aminata

Group C2

PROJECT REPORT

For this project, my classmate and I aimed to make a facial recognition system using
Matlab networks.. We had to build a face recognition system with the following
design specification: able to recognize the frontal view faces and Detection (and
registration) of the face in a test image must be automatic.

From the beginning of the project, we decided to do, both of us, all the steps of the
beginning as we do the project in distance, it was not too obvious then each one
concentrated on different networks, for me it was the resnet50 and resnet101. So first,
I saved all the images from moodle to create a training folder and a validation folder

Then as indicated, I rotated the images before the training, so that they are in the right
direction.
Khoule Aminata
Group C2
Then in a second step, I used a code moveToFolder.m to create a folder for each
student.

After all these steps, I concentrated on training the networks. First, I installed the
RESNET50 and replaced the base layers with myown layers from the menu and
connect to network and changed numFilters in convolutional2dlayer to the number of
classes in my data. After importing the data, I started the training which lasted 6 hours.
Unfortunately, after this training, my computer did not want to do any more training.
Khoule Aminata
Group C2

Then, I had a facial recognition of 66.39%.

Finally, I was able to do the training with the inceptionv3 network with which I had a
facial recognition of 70.0820 %. I had to change the recogniseDeep code and put
[299 299] instead of [244 244].
Khoule Aminata
Group C2

To conclude, this project was a very rewarding project because it allowed us to


practice what we learned and to fully understand the functions of artificial intelligence.

You might also like