Image Classification

Is the table clean?

Yes, that was the name of my final year project in polytechnic. An interesting one indeed, it was an experimental project on image classification technologies. My team and I work on two objectives for the project. The main focus of the project was to use image processing tools to assist in the checking of tables to determine if the tables are clean. The use cases for this application will be for food centres and hawker centres to automatically dispatch cleaners if there is a detection that the table is unoccupied and dirty. A secondary objective is to measure an object's dimension (such as dimension of a newly painted wall) by using a reference object (such as ATM card). This will help in the automation of job billing and measurement without the use of a measurement device

Technologies used:

• Android

• OpenCV

• Clarifai API

• RScript

• Python

Data training process

Training of data set is done via OpenCV and the Clarifai API. Images are first pre-processed via Python and OpenCV. Techniques such as changing of colour channels and canny filters are applied, this is useful for increasing accuracy for training and testing the dataset and for better detections of edges (used for for measurements).

Final Application

All the datasets and technologies and placed together into an android application to showcase the potential uses cases and implementations. The application is able to pre-process images, and add images to the dataset. Additionally, the application can measure objects using a reference object (any standard card), identify and tag items (to detect if the table is clear) and to detect other objects (like handwriting numbers, faces, etc).

Read More

https://opencv.org/

https://www.clarifai.com/