README.md

MATLAB_RockPaperScissors

MATLAB Deep Learning Session Files
Link to matlab drive: https://drive.matlab.com/sharing/e3f8c4c7-6a26-4f2b-9a2b-1193de247686/

Beginner's MATLAB Intro to Deep Learning Session.

Make sure you have a MATLAB Login! Use your University Email to activate!!!

Instructions:

1. First open the files in MATLAB Online (or local if you have it installed)

2. Select "Copy Folder"

3. [WEBCAM] Open collect_images_via_webcam.m and run

[NO WEBCAM] Take pictures of ROCK, PAPER, SCISSORS, NULL class with your mobile phone. Upload images to ./data/CLASS

4. Collect 20-30 Images of each class. Move your hands in different areas.

5. Open transfer_learning.mlx

6. Seleect "APPS", then "Deep Network Designer" in the dropdown menu

7. Select googlenet in the Deep Network Designer

8. Delete loss-3-classifier layer

9. Replace with fullyConnectedLayer

10. Change Output Class to 4

11. Unlock output layer and select export

12. lgraph_1 should be visible in the bottom right. Select Run in live editor

13. Training will start.

The accuracy should increase whilst loss decreases. Close when training has been complete.

14. Test network!

We can test the image on images taken on any device! Try an image taken with a webcam/phone, upload to PC and run the following:

inferImage(netTransfer, 'PATH_TO_YOUR_IMAGE')

15. Run on Webcam

In the MATLAB Command Window, run the following:
run_network_on_webcam(netTransfer, 0.4)

Now we should get this.