DoggoRe

Ellie

Lost dogs don't have to stay lost forever. DoggoRe is an Android application that helps recognize lost dogs with the help of a convolutional neural network, using transfer learning with Inception-V3. It stores missing and found dogs in a database and uses that to display similar dogs to a user who lost a dog or found a missing dog.

DoggoRe uses a convolutional neural network trained in Python using Keras with a TensorFlow backend. TensorFlow Lite is used to use the CNN on the app. The app also uses OkHTTP to call PHP scripts on a server which access a MySQL database. This database holds the information of missing and lost dogs.

Data Flow

Activity Flow

Projected Timeline

2/17 - Pull data from server & popular scrolling lists in Android app

2/24 - Be able to click on an item from those lists to bring up a new activity

2/24 - Have first neural net trained/ in the process of training

3/3 - Have first neural net working in app

3/17 - Inferences can be made using NN (includes getting photos from phone, although could be pushed back and use project assets instead)

3/31 - Next version of NN in app - more intricate analysis

4/7 - Emails sent out accordingly; app looks cleaner

4/14 - Notification when new missing dog is added

4/21 - Last NN; other measures to narrow it down possible (maybe even a second neural net or a compilation with other ML algorithms)

Likely/ Possible Delays

Server usage - no idea on that
NN training time
Losing days of time to grad school visits, probably all in March

Gantt Chart

Gantt Chart

This website was made with the help of Bootstrap.