Chris Pollett > Students >

    ( Print View)




    [Deliverable 1: GPS App]

    [Deliverable 2: Image Feature Selection]

    [Deliverable 3: Accelerometer and Video Recording Apps]

    [Deliverable 4: Model Survey]

    [Deliverable 5: CS297 Report]

    [Menze Paper Review Slides]

    [Pomerleau Paper Review Slides]



Feature Selection on Discrete Image Collection


Feature selection is a vital part of machine learning and deep learning. By investigating how each feature impacts the overall performance of a model, we can differentiate the features based on their discriminatory powers and therefore focus on those more impactful features. The model will have less dimensionalities and complexities as result, and will consume less computing power and resources as well.

How to run the code:

Two ways: emulator or android smartphone tethering. I mainly used emulator and double checked with tethering. Both work well albeit slower implementation on emulator.

For emulator, choose any virtual cellphone that supports Android API 28 and click RUN on Android Studio. Emulator provides the GNSS coordinates: click on ... button and GPS information is listed on the top of left-handed bar.

For cellphone tethering, a Huawei MateSE is used for testing. To debug and run the code on the cellphone, two steps are required: 1)activating developer mode on the cellphone. Follow Setting > About phone > Build Number > click seven times, and you are in developer mode. To allow for debugging and testing, follow Setting > System > Developer Mode > Debugging > select USB debugging; 2) choose the tethering mode in Android Studio by following RUN > Edit Configurations... > Delopyment Target Options > USB Device.

Source code: