AutiSpot is aimed at being one integrated stop for the autistic community. A cumulative platform to detect Autism via Machine Learning & Neural Networks also providing several functionalities such as; personalized communication boards to help develop cognitive abilities in children, a Social networking platform to connect the 'ASD'.Autism Spectrum Disorder community and much more.

As a research assistant, at Muffakham College of Engineering & Technology, I worked under the guidance of professor Dr. Syed Shabbeer Ahmad to conduct a research on detecting ASD in toddlers before it's onset, and curating innovative approaches to improve their cognitive performance using Neural Networks and Deep Learning. We completed the study, developed a prototype and presented the research at the 2022 Second International Conference on Artificial Intelligence and Smart Energy (ICAIS), now published at IEEE. We were awarded the Most Innovative Major Project of the CS Dept. MJCET for our contributions.

View Paper

• View Implementation
• View Demo - Youtube

• Key-Feature Breakdown:
  
  ◦ Autism Detection Using Machine Learning
  ◦ Autism Prognosis using Neural Networks:
  ◦ Personalized Communication Boards
  ◦ Automated Todos
  ◦ Social Networking Platform
  

• Technologies Used:
  ◦ JavScipt, Python
  ◦ MySQL, Flask
  ◦ NLP, OpenCV, Convolutional Neural Networks (VGG19) & Supervised Learning Techniques (Random Forest, Decision Trees etc.)
  ◦ HTML, CSS, AngularJS

Autism Detection Using ML & CNN
Upon performing exploratory data analysis and feature engineering on raw data, three supervised machine learning algorithms namely logistic regression, decision tree and k nearest neighbour are used to classify data for diagnosis. The users are required to fill a checklist consisting of a set of questions based on behavioural, health, and hereditary parameters. A report is generated to display the results based on the predictions.
Additionally, The VGG19 model is imported from tensorflow keras in addition to the pre-trained weights on the ImageNet dataset to perform feature extraction, fine-tuning and predictions on the provided dataset. The hyperparameter (epochs) that explicates the number of occurrences that the learning algorithm runs through the dataset is performed in ten iterations. The system expects the end user to upload an image and based on the analysis provided by the CNN model, the system projects the result of the person being autistic or not.

Personalized Communication Boards
The module that lets users select boards, hear sounds and words pronounced aloud that can help them verbally communicate their needs. This kind of distinction of words and frequently communicated gestures are provided through this module which helps the autistic people overcome their most noticeable barrier of communication. This module also provides the functionality to utilize the personalized communication boards provided with filters to help them convey their thoughts by simply tapping on the cards to generate sentences. In order to generate the speech, synthesis service which regulates the creation of speech using textual content provided, the following interfaces of the Web Speech API is used.
Some of the methods used include: speak(), pause(), resume(), and getVoices()

Automated Todos
The primary objective associated with this module is to facilitate users via an intuitive user interface to organize, prioritize daily tasks and reinforce them by enhancing these abilities. The system is devised using flask and a MYSQL database to store tasks, and users are provided with a list of frequently performed activities. The users can also add tasks, mark them as complete or incomplete as well as view the total tasks to be performed in a day, ones that have been performed and tasks yet to be completed. The tasks can also be deleted once no longer required or can be recursively used just by updating the status of the to-do for the day. In order to make the tasks visually comprehensible, images and other interactive features are included.

Social Networking Platform
People in the autism spectrum often feel overwhelmed during interactions. This feature emphasises on facilitating them by augmenting their participation in the community, enhancing their social skills and by providing a platform to display their talents. The system is developed using the SQLAlchemy Toolkit extension provided by Flask which is an object relational mapper that maps object parameters to the underlying database table structure and converts function calls to SQL statements implicitly. Features include add posts, like/dislike posts, bookmark, comment, view later, search posts by name or tags, create personal profiles, etc.