Talking with animals

Brief from Posthuman perspective

The project is for demonstrating the future state of the sightseeing vehicle, which is not only a vehicle but a bridge for connecting human and animals, In a short way, using AI to detect what the animal is, then teach travelers to learn about their habit and emotion. Based on ML we could translate the animal’s language to human language. 

So all the creatures’ voices on earth could be heard. the ecosystem between human and animals is not just the awareness of animal protection, but try to communicate with animals in a comfortable environment. 

The ecosystem between human and animals is not just the awareness of animal protection, but try to communicate with animals in a comfortable environment. From a speculative perspective, it helps human to deeply understand the behavior and emotion of wildlife animals. Humans and animals wouldn’t have boundaries anymore. Humans coexist with animals, they learn from each other and grow with each other, Autonomous car could calculate the safe distance apart from animals, to provide the best condition and assist with communication. Let us solve all the prejudice and build the world a better place.

Poster

Process

Exploration

At first I was thinking of using the runway ML(coco-ssd) for making the animal facial recognition, but after a try I found that runway has it’s own database and other people are not allowed to make changes, it is obvious to see from the right photo that it accuracy is suck cannot detect what animals is, Also I tried the wekinator but the OSC input are extremely hard to train and a lot of key parts were missing.

Steps

So I end by using the teachable machine which is a a web-based tool that makes creating machine learning models fast, easy, and accessible to everyone. I started with using the web came taking pictures of 10 types of animals about 200 images for each type. Then train the models through teachable machine, then export the model to the ps js. After that I added the sound files for each animals, the goal is when the camera detect the animal then the sound will be played which imitate the conversation between human and animals.

sample sound

TEACHABLE MACHINE

Materials

CHALLENGE

The challenge I encountered with is the teachable machine need to have a clean ground for better recognation, so I re-took the pictures from a white background. the second challenge I have is uploading the sound to p5js and let it play when animals get detected. Big thanks to lee for helping me figure out the code. Third is the arduino elegoo robot cart. which I bought the materials online then assemble it and upload the load the codes to its phone app so that achieved the bluetooth control through mobile phone. The cart it selves has line tracking function which similar to the autonomous vehicle. The reason why I prefer the autonomous vehicle is the near future autonomous will replace the human-drive car. and it contains more developed technology and more efficient to help finding and communicating with animals.

NEXT STEP

For the next step I would consider that how to delay the sound files to make 3-5 second delay for each animals so that the sound won’t be overlapped. The accuracy of the robot cart can get improved it should stop when detect animals, Also imitate the human and animal talking scenario.

AND Special Thanks to Christine, Lee and all the guest speakers for supporting me finish the project.