- When we envision the skyline in the game Cyberpunk 2077, we realise such a future is not what any one of us expects. Immense energy consumption is the top one reason that leads to such a future, and thus we want to do something right now. 
- We interviewed 40 of our friends and schoolmates about their daily issues on energy waste, 33 of whom found forgetting to turn off AC was a substantial problem especially when they are busy. And 28 of those believe this could lead to massive energy waste.
- Therefore, to save our planet and environment, we decided to start from the basic but practical idea--automatic AC controller.

What it does

- The auto controller detects and senses the number of people, humidity, and temperature in a room. 
- As long as there exists at least one person and, for example, the current temperature is higher than the setting temperature inputted from the front end, the auto controller will automatically turn on the AC. And when no one is detected by the controller, the AC will be turned off, which helps us save lots of energy. 
- Users can also monitor the status of the AC, temperature, and humidity in the room from the front end, so that they can better manage their own way to live an environmental friendly life.

How we built it

- Visionseed from Tencent is used as the camera and the place to deploy our independently trained person detection model (ResNet pretrained on ImageNet and trained on CrowdHuman). 
- The detection signal is transmitted through physical connection to Raspberry Pi. And Raspberry Pi utilizes an augmented accessory DHT22 as the temperature and humidity sensor, and combines those signals to send to the server using Socket package in Python.The server (Google Cloud) receives the signal and updates the room temperature, humidity, and AC status which will be presented on our fancy HTML5 page through the web framework Flask in Python. The front end can also send the set temperature to the server and feedback the order of whether turning on or off the AC to the Raspberry Pi, which will send the signal to the AC by Infrared protocol.

Challenges we ran into

- The Raspberry Pi breaks down at the middle of the hack, and we can not longer get the temperature and humidity data
- Although our programs run smoothly at local servers, we fail to transmit local data to Google Cloud. 
- Our teams members are working from different cities in China, so it takes us some time to  figure out an efficient way to collaborate remotely. 

Accomplishments that we're proud of

- The machine learning algorithm works out pretty well at detecting moving and still people, and thus it yields accurate results of the number of people.
- We've learned to use some new languages, modules, and packages in two days, which are challenging but fun. 

A decent website is built for introducing our project and for future users to control their ACs

What we learned

- Gain a deeper understanding of IoT, machine learning algorithm, and how does a website work.

What's next for Autoconditioning

- We have to spend more time learning to use Google Cloud and develop an effective way of transferring real-time data from local hardware to cloud server and reflecting changes in data on the website. 

Built With

Share this project: