Rent a Chicken

Initial problem

The standards of chicken husbandry for the production of eggs are mostly non-transparent and not animal-friendly. The appreciation that should be given to the chickens is often insufficient. Moreover, there is no adequate alternative to eggs and few people have the know-how or the possibility to keep their own chickens.

Solution

Rent a Chicken understands chickens better than anyone else. With our innovative AI, we can communicate with our chickens based on their clucking and thus determine whether they have just laid an egg. The first chicken interpreter!
With the help of our AI and an accompanying app, it should be possible for every animal-friendly chicken farmer to sell eggs simply and easily. Users can rent the chickens via our app, see them live, view the current egg status and pick up the eggs.
With Rent a Chicken, we want to promote regionality and a sense of community in places with participating chicken farmers. But we were mainly driven by the idea of developing a concept in which the chickens are kept in a species-appropriate and loving way. Users can bring a piece of nature and a farm directly into their own homes via their mobile phones and thus contribute to more sustainability and more appreciation for chickens and chicken farming.

As part of the Prototypes and Redesign lecture, we developed Rent a Chicken, the first functional chicken interpreter.

Process

Moodboards

We started here with mood boards to get a feel for the theme and used these to develop the colours for our design.

This resulted in the following colours for our design:

Here is a small selection of Mid Fidelity variants as an example:

Mid Fidelity

Here is a small selection of high fidelity variants:

High Fidelity

Stack

Here is a small selection of low fidelity variants:

Low Fidelity

Our frontend is divided into three areas. On the main page you will find the egg counter, where users can see how many eggs are currently ready for collection and can confirm the collection process via the integrated button "collect". An overview of the rented chickens and other chickens on the farm can also be found on the start page. In addition, a live cam transmission is available so that users can experience their chickens authentically. Another important point on the home page is the profile view.

There, the tenants can change their payment method, cancel or even change the farm.
Another part of the app is the "Knowledge" section where users can learn new things about their rented animal friends at any time. The last section includes an alarm clock function. The tenants can be woken up live every morning by the rooster or roosters of their farm, in order to be even more closely connected to farm life. Optionally, an individual time can also be set

Main View

Item Two

Change the text to include your own content. Adjust the font, size or scale to customize the style.

Item Three

Change the text to include your own content. Adjust the font, size or scale to customize the style.

Raspberry Pi

The Raspberry Pi, which is extended by a camera and SIM stick, is located directly at the scene of the action, namely the chicken coop. Audio recordings of the background noise in the chicken coop are made and saved via the camera. Recordings are only made if a certain value for volume and degree of movement is exceeded. This is set in a motion detection script on the Raspberry Pi. An audio recording always lasts 10 seconds. Another script on the Raspberry detects that a new audio recording is being saved and runs it through the AI. Our AI then analyses the files saved in WAV format. If it recognises the unique clucking sound that chickens make after they have laid an egg, a one is sent to the backend via MQTT.

After the AI has learned, it is stored in a model. Then it can be addressed by various Python scripts and gives us its prediction as an answer.

In our case, the AI uses 10 spectrograms to evaluate whether an egg has been laid or not.

AI Snippets

AI

The project is based on an AI that works with the help of Python and TensorFlow. The Tensorflow database has a ready-made audio AI that you can train with Python and integrate your own data.

How does the AI work ?

As described above, the AI is fed with various audio files. To develop the AI, we initially recorded these ourselves in the barn, edited them together and categorised them according to ordinary clucking and clucking after an egg has been laid. In order for the AI to be able to process the audio recordings, a Python script has to cut the audio recordings into 1 second long audio recordings and then turn them into spectrograms.

The AI then has a large data set at its disposal with which it can learn and predict. Our model learns in 50 epochs, and in each epoch all the audio files are gone through and predicted and learned.