Fruit-Checker

Inspiration

The fact that a fruit quality checking tool can actually be employed in real life inspired me to do the project.It is quite beneficial in the sense that it can save time and manpower, help in building quality food products by identifying fresh produce and prevent other fruits from ripening faster by early detection of spoilt fruits.A lot of automation is happening these days, so building something like this was really interesting.Also, hadn't used SashiDo nor google's teachable machine before, so this was a good opportunity for me to learn about them.

What it does

It is a web application that can be used to check if an orange,apple or banana is rotten, so basically it is a fruit quality detection tool.Users can either use their webcam to get live predictions or upload an image and get results.

How I built it

The machine learning model was trained by feeding image samples of fresh and rotten fruits using google's teachable machine.100 images were taken from each class.As it requires a small dataset only 600 images were taken in total.Javascript was used to add functionilty to the webpage using which images can be uploaded or prediction can be done in a continuous way using a webcamera.Using the trained model, we can obtain probabilities of the image belong to each of the classes, from which the highest case is identified and the respective class name is shown.Currently , image uploads are saved in the server.I realised that it wasn't necessary to upload files to the server but it could be useful to gain more samples from the public and get a better dataset

Challenges I ran into

It took me a while to figure out how to upload files to the parse server and then use it for prediction.The files would get uploaded but i would get the same prediction result each time for different images,I wasn't sure why that was happening.I tried using a canvas element to display the image instead of directly using the image, and it worked.I did not find a dataset which includes various other fruits so it is limited to just 3 classes for now.

Accomplishments that I am proud of

I was able to built a cool web application(atleast i think so) with all the functionalities mentioned and i'm happy about it.This is my first time participating in a hackathon solo so i had to take care of the entire application by myself and it was a great learning experience.

What I learned

I learnt about various aspects of SashiDo and Google's teachable machine and improved my javascript knowledge too.

What's next for Fruit-Checker

I haven't really found a dataset for vegetable images.I'm keen to do a similar thing with it and also add more classes to the existing project.I am also thinking builing models using python by taking more samples and learn about various ways in which it can be accomplished.

Built With

• css
• googleteachablemachine
• html
• javascript
• sashido