IKSHANA

The idea is to create a device that detects the objects and gives audio output in real-time. The following steps happen in the loop of detecting objects: taking video input, detection, audio output.

Comment 2

Inspiration

We got inspired by watching visually impaired persons around us.

What it does

This project enables visually impaired people to detect objects without actually touching them. It consists of a camera module and a speaker module. The camera module captures the video stream in real-time. The captured frames are then processed and when an object has been detected the name of the object is given as output.

How we built it

We built our model using a raspberry pi zero w. We used a pi camera for capturing video and the output is given using the speaker module. The image processing model is trained using images from google. Our object detection detects 90class objects. Open cv and pyttsx3 modules are used.

Challenges we ran into

Our first challenge was Budget. Other versions of Raspberry pi like 3,4 are expensive. so we opted for a zero version. The next challenge we faced was we thought of implementing it using TensorFlow but couldn't since the packages were taking hours and hours to get downloaded so we used cv2. The last challenge we faced was we weren't getting the output as audio from the audio amplifier module so we used a Bluetooth speaker.

Accomplishments that we're proud of

We successfully implemented object detection using a raspberry pi zero and gained experience in computer vision.

What we learned

We got to know about the raspberry pi, computer vision, and python libraries such as os and cv2. This is our initial attempt at creating a project under computer vision using raspberry pi. We learned about much technical stuff such as datasets and got a deeper understanding of python. We even got to know about Anaconda, Jupyter Notebook.

What's next for Ikshana

Technology is never constant. With the updating technology along with modified software and hardware, our device can be made to work in an expedited and much more efficient manner. We need to determine the object’s range from the person. We are planning to use a tensor flow API. Our aim is to make the device that is portable, can be carried anywhere, and make the model to be personalized to the user.

Built With

Share this project:

Updates