Aimify

An archery scoring app designed to end the use of paper score sheets by offering a modern and convenient alternative using Computer Vision.

Planned Features

• Sleek and modern UI made to Material V3 guidelines

• Automatic scoring via the use of instance segmentation models

• Cross platform compatability through the use of React Native

• User interaction and sharing

• View all of your earned badges

Machine Learning

Figure 1: Images being labeled in CVAT, a free and open source dataset labeling tool.

Machine Learning will be used throughout Aimify to facilitate a key feature: automatic scoring.

Through a YoloV8 model trained on a dataset consisting of over 200 images of archery targets (currently in production), users can calculate their scores by simply holding their phone up to the target and pressing a single button.

The trained YoloV8 model will then be exported in the Tensorflow lite (.tflite) format, enabling on-device inferencing for an offline experience.

Current Progress

Figure 2: Current progress on arrowV1 model

The above image shows the first iteration of arrowV1. This version uses Roboflow's 3.0 instance segmentation model and is a work in progress. Using this model comes with some drawbacks, notably no easy way to download the model, and less than satisfactory result, causing approximatly 50% of the outer rings to not be masked. This can be fixed through using YoloV8 models and increasing the number of images in the dataset respectivly.

User Interface (In the backlog)

Figure 3: UI mockups created in Figma with the use of Material V3 components from Google.

The majority of archery scoring apps available today use dated UI designs, either from Material V2, or earlier. To combat this, Aimify will use Material V3 (Material you) theming and components

Aimify will be built using the React Native framework via the Expo implementation, allowing for a single codebase to be deployed to both Android and IOS. In addtion, Aimify will use React Native Paper, giving acsess to hundreds of Material V3 components to build the UI from.

Roadmap

• Finish annotation of dataset (Completed 31/05/2024)

• Train ML model by end of June 2024 (Ongoing)

• Start UI by July 2024

• Work on backend and server-side software by August 2024

• Release alpha build by September 2024

Though this may be subject to change in the future, the current Aimify license is based on the GPLv3 license, with an added restriction preventing the software from being monetised.

A PDF-formatted version of this license can be found here.