Report

The report is in the REPORT directory.

Installation

Start the terminal and you can install the virtual environment as follow:

# Go to assignment folder
cd ML395_NN

# Change permissions to execute bash script
chmod 755 install_lab_env.sh

# Create virtual environment 'env' and install requirements
# This can take several minutes to finish
./install_lab_env.sh

How to run src/test.py for Q5 and 6

Here is the example of how to run src/test.py.

model_path_q5 = "PATH/TO/THIS/REPO/pkl/final_model.pkl"
img_folder_q5 = "PATH/TO/IMAGE"

# Call test_fer_model() for Q5
test_fer_model(img_folder, model_path)


model_path_q6 = "PATH/TO/THIS/REPO/src/question6/models/vgg_netvgg.hdf5"
img_folder_q6 = "PATH/TO/IMAGE"

# Call test_deep_fer_model() for Q6
test_deep_fer_model(img_folder, model_path)

train_fcnet_optimise_params.py is used to train the optimised model for question 5. The model can be trained directly from the JPEG data on bitbucket, however this is slow to load. Also provided is a function to load the data from a pickle object, which loads significantly faster.

train_fcnet_optimise_params_gridsearch and train_fcnet_optimise_params_gridsearch_2dims are used only to search for parameters and were not used to train the final model.

Instruction

395 Machine Learning: Assignment 2

Please notice that this year, we have an intermediate and an advanced version for the second assignment. Identify if your group consists of:

1. Master's Computing students: You must complete this advanced assignment (manuals/assignment2_advanced.md, pdf format is also available).
2. Non Master's Computing students (Bachelor's Computing/non-Computing/external/exchange): We recommend you complete the the intermediate assignment which is available on the group web site for download. Optionally: You can also choose this advanced assignment, it is up to you.
3. Mixed of Master's Computing and non Master's Computing students: You must complete the advanced assignment (manuals/assignment2_advanced.md, pdf format is also available).