• - Understand how kivy renders UI
• - Learn about .kv style sheets and how to stylize UI
• - Try to replicate UI Sketches

Draw wireframe sketches of the UI for each screen a user will see when using the app.

• Draw sketch of the start screen the user will see when launching the app
• Draw sketch of the photo capture screen
• Draw sketch of the pokemon statistics screen that will display information

Add sprint review views to board. Explained here.

• Francisco
• Geovanni
• John
• Yousef
• Elliot

Test an object detection model for detecting Pokémon in an image.

• For each Pokémon the algorithm has been trained for, take 5 different pictures of a Pokémon and record how often the app identifies it

Use pretrained object detection model to detect Pokémon in an image.

For each pokemon the algorithm has been trained for:

• Take 5 different pictures of a pokemon and record how often the app identifies it

Create scripts that will prepare the dataset for training the Yolo algorithm.
Script functionality:

setup.py

• Create images and annotations directories
• Put image and annotation files in their respective directories

prepare_training.py

• Extract data from XML annotation files
• Convert data to Yolov5 format
• Store data in respective TXT files
• Split the dataset into training, validation, and testing images
• Move images to their respective directories

Attempt to train the Yolov5 model using the provided sample road sign dataset that is known to work.
This is to ensure that the training works correctly prior to being performed on the final Pokémon dataset.

• Add yaml file road_sign_data.yaml to specify location of image directories and labels
• Edit the configurations file for the yolov5 training script hyp.scratch.yaml to reflect desired parameters
• Rename annotations directory to labels
• Run the yolov5 training algorithm train.py on the prepared road sign dataset (requires completed #26)
• Evaluate the training results and ensure they match those of the tutorial
• Test the resulting model on the provided test images using the testing script

Complete the research required for understanding how to implement the object detection.

• Determine which version of Yolo to use
• Find a good tutorial for training a Yolo model
• Find a good tutorial for using a Yolo model with pre-trained weights

Review peer feedback for our team.

• Access student feedback on blackboard
• Go through comments made by each student.

• Update colors of types for legibility

Find images of Pokémon to train the object detection algorithm.

• Locate a publicly available database online
• Import the database to your local machine
• "Prune" or remove images from the database so that it only includes the select gen 1 Pokémon to use for testing the model
• Gather additional test images that contain multiple Pokémon in each image

• Put various datasets of pokemon into one consecutive dataset
• Make sure labels are approperiate such that 2 datasets don't overlap
• Make sure the classes are named properly such that a class does not have different names for the same pokemon

• Write a function to assign a color to a passed list of types
• Name
• Picture
• Description
• Types
• Weaknesses
• Abilities
• HP
• Attack
• Defense
• Sp. Attack
• Sp. Defense
• Speed
• Moveset
• Evolutions

Create a more streamlined process for training the Yolov5 model. The current process requires renaming the annotations folder to labels and executing multiple scripts in the correct order with the correct command line arguments, which is not ideal.

• Combine setup.py and prepare_training.py scripts
• Modify test.py to use "labels" directory rather than "annotations" directory → upon further investigation, the tutorial calls it test.py, but actual script it's referring to is val.py
• Write out commands and arguments (requires #31)
• Debug errors

Create a basic app that displays statistics on a specified Pokémon, as obtained from an internal database (csv file?).
This information will display in a new screen after the user has taken an image and object detection has been completed.

• Name of Pokémon along with a description of each one.{Geo}
• Type weakness(es){Geo}
• Competitive information (move list, type of team member, recommended held item){Geo}

Train algorithm on Pokémon object detection dataset.

• Prepare images and annotations
• Run the training script

• Download Kivy Launcher App
• Get Basic App Running From App
• Get Camera App Running

• Research the different Yolov5 architectures to determine which would be best suited for this application.
• Research the Yolov5 hyperparameter options to determine how the parameters will affect the model.
• Determine the hyperparameters best suited for this application.

• Gather json data for evolution and weaknesses
• Python function that returns this information when passed an ID

Understand how to access Pokémon information and download it locally

• Access API Files for a Pokémon to use it in python

Following this tutorial, learn how to automatically optimize the training hyperparameters.

• Make required modifications to Yolov5 code
• Run hyperparameter evolution on the test road sign dataset

Build a Home Screen using Kivy for the app to display once the user starts the app.

• Display the App name at the top
• Display instructions to use the app
• Include a Button to open the camera
• Display a list of pokemon the user can select

• Stats
• type
• ability
• move set
• code for remaining 151 Pokémon

Create working functions to separate each Pokémon's information for the first 25 Pokémon

• Stats
• type
• moveset
• ability

Create an object detection model for detecting Pokémon in an image.

• Develop a Yolo object detection testing algorithm - John

Use the setup.py and prepare_training.py scripts from #25 to prepare the sample road sign dataset.
This is in order to test the Yolov5 training algorithm on this known-to-work dataset first to ensure no errors prior to using it on the final Pokémon image dataset.

• Download road sign dataset
• Move images and annotations into the base working directory
• Run setup.py
• Run prepare_training.py

Installing and getting familiar with Kivy.

• Install Python 3
• Create a virtual environment in Python 3 (optional)
• Install Kivy using the pip package manager
• Install TensorFlow for ML development
• Ensure installation works by running a basic hello world script that imports Kivy and TensorFlow

Create a basic app that opens the phone camera, takes a picture, and saves the picture on the device.

• Make start screen in Python
• Add button to capture photo
• Create photo capture function in Python
• Save captured photo to phone

• Get an Android-x86 VM running on PC
• Build app and transfer it to VM
• Try to run the app
• Take note of any screen scalability issues and think of ways to solve them

Research how to use a trained yolo model to identify objects in new images

• Find out if the yolo training algorithm must be included in app
• Find out package requirements for using trained model

Attempt to run a sample yolo model included in the yolov5 package

• Load yolov5s model
• Obtain and print results from yolov5s model

• extract and gather information for each Pokémon

Train an object detection model for detecting Pokémon in an image.
This requires the #8 and #2 Tickets to be finished

• Train the algorithm on the training data images - Francisco