Squatty Bird

• 

  Squatty Bird

• 

  It's a bird!

During this quarantine season, Richard and I have become so lazy at home and not exercising much. Thus, we came out with an idea to make us exercise! User controls squatty bird by squatting and standing. When the user holds a squat position, the bird doesn't do anything and when the user stands up, the bird jumps.

We built the game using pygame. As for the image classification model, we built it using tensorflow and opencv. There were definitely a few challenges that we ran into. It's hard to find proper squatting and standing images. Most standing images on Google are showing babies. This will cause a bias in our neural network. Thus, we used UCF101 dataset to obtain a bunch of videos. Then, we converted the videos into images. We labelled around 14000 images ourselves between squatting and standing. Another challenge is this was a last-minute idea. We only started building it last night so we had to rush through it.

Although we did this last-minute, we got the game running properly. It was so satisfying to see it running properly. We are hoping to add more exercises to Squatty Bird so the user can control the bird with more actions.

Built With

• opencv
• pygame
• python
• tensorflow