Food Critic-Texture Classification

This mobile application aims to provide quality of a food by looking at its texture.It uses Haralicks GLCM Texture Classification method to provide quality of recipe/food/agricultural commodity.

Comment

Inspiration

I've seen apps for ordering food online or to find recipes, I haven't seen one to find the texture of the prepared food.It inspired me to develop this app using AI Modeling techniques. The app has been tested on real android device 11+ and the output is displayed as demonstrated in the demo.

What it does

Provide Texture classification of food by choosing an image from photo gallery or camera. Know the texture in One-Click. Model is currently trained on well prepared cake recipe.

How we built it

With python programming developed a grey texture image using GLCM Haralick Classification. Used the texture to create a TensorFlow Keras Sequential classifier and trained a model with loss/accuracy of 5/5 - 0s - ** loss: 0.1041 - accuracy: 1.0000 **- 219ms/epoch - 44ms/step

Challenges we ran into

Flutter/Dart does not provide plugin to use the python mahotas library to produce a grey image using haralicks feature.It was challenging to figure the backend server to get the texture back to the app's frontend programming to compute the result.The texture and the probability calculation from server to client was achieved using the flask server on python programming.

Accomplishments that we're proud of

Was able to predict the texture of the chosen image.Proud to know that this texture classification and prediction can not only be achieved on food texture but can be applied to different areas -medical, environment

What we learned

Learnt modeling techniques.Learnt how spatial difference help in classification of nearly identical images.link And new ways to solve problems when functionality didn't work.

What's next for Food Critic-Texture Classification

Train the model on different datasets to achieve the result. Introduce categories (medical, environment) to accommodate other areas. The subject maybe considered for not finding the texture of food alone but also for a broad range of topics -medical scanning of images, -environment/climate change detection

Refactor the model achieve to accurate prediction

Built With

Share this project:

Updates