Indoor Object Identification Using Machine Learning

Project Title

Indoor Object Identification Using Machine Learning

Project Area of Specialization Artificial IntelligenceProject Summary

A prototype device will be developed using a processor and camera to predict the objects and their proximity from the blind user in indoor settings. For object classification, TensorFlow object detection API and a customized training model will be employed. The device will process voice inputs from the blind user for identifying the object as dictated by the user. For feedback and speech recognition, headset can be used. While, ultrasonic sensor can be used for proximity or distance measurement. This device will be embedded on blind person’s cane.

Project Objectives

• To target Sustainable Development Goal 3 (Good Health and well-being), Goal 9(Industry, Innovation and Infrastructure) and Goal 10 (Reduced Inequalities).
• To improve independence and quality of life for blind and partially sighted people in an indoor environment.
• To easily commercialize it to blind community (institutions and schools).
• According to WHO, the number of blind people is predicted to rise from 36 million to 115 million by 2050. This device will be a great support tool to assist blind in their day to day activities.

Project Implementation Method

• This project is based on Machine learning framework, TensorFlow.
• A customized training model will be employed on the processor. 
• A dataset of 200 images will be taken for each indoor object.
• More images will lead to more accuracy of prediction. 
• Faster identification of objects and frame rate depends on the speed of processor, hence processor having atleast 8 GB RAM is needed.
• Webcam will capture the live video stream then processor will predict the captured objects.
• A high quality webcam is needed to capture the objects clearly.
• To give auditroy feedback to the blind, headset will be used. It will also process voice commands of blind user.
• Distance and proximity of objects will be measured through ultrasonic sensor.
• Cables, board case and setups will be used to integrate all the components.
• A blind person's steel cane will be used to embedd this system on.

Benefits of the Project

• This project will achieve SDG 3,9 and 10.
• This project will assist blind people at workplace and home.
• People with vision disabilities are handicapped in perceiving and understanding the physical reality of the environment around. with rapid advancement in deep learning, smart cameras, object detection techniques and portable computers, it is possible to assist blind people by designing a wearable aid which alerts users by the presence of obstacles and provide guidance to grasp their desired objects.
• Blind and visually impaired people can easily carry this system, since it will be embedded on a single cane.
• Independence and quality of life for blind and partially sighted people in an indoor environment will be improved.
• This project can be further extended by adding more features of communication and advanced alerting functionalities.
• According to WHO, the number of blind people is predicted to rise from 36 million to 115 million by 2050. This device will be a great support tool to assist blind in their day to day activities.

Technical Details of Final Deliverable

• A cane with system embedded on it.
• connected with bluetooth headsets.

• This device will provide following features:

–Real-time object detection and Classification using customized model.

–Proximity of objects

–Auditory Feedback

–Processing of Voice Commands

–Audio feedback when target object is found.

• A document or manual with described instructions to be followed in order to use this system.

Final Deliverable of the Project HW/SW integrated systemType of Industry IT , Others Technologies Artificial Intelligence(AI), Internet of Things (IoT)Sustainable Development Goals Good Health and Well-Being for People, Industry, Innovation and Infrastructure, Reduced InequalityRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	79000
SmartFly info HiKey 960 Single Board Computer	Equipment	1	53000	53000
Logitech C922x Pro Stream Webcam	Equipment	1	13000	13000
Blind person’s hard, steel cane	Miscellaneous	1	5000	5000
High quality Bluetooth Wireless Headset	Equipment	1	2000	2000
Ultra sonic sensor	Equipment	1	300	300
Board Case	Miscellaneous	1	1000	1000
Cables	Equipment	1	1000	1000
Power bank	Miscellaneous	1	3000	3000
button board	Equipment	1	700	700