IKSHANA- AN AI APP FOR THE BLIND
The evolution in the field of artificial intelligence has caused an increase in the development and usage of numerous virtual assistances. Some few popular ones are Siri, Alexa, Google Assistant, Microsoft Cortana, etc. Thanks to the invention of these AI assistants, everyone can now have a personal assistant with them. So now you can just call out your virtual assistant to set the time for your date tonight without forgetting it. Again, for the love of god. In other words, we can carry out our daily tasks without putting in much effort, interlink our devices and automate tasks without the human touch. It is true that technology can make lives easier and convenient. But it can also make some people’s lives difficult. The visually impaired, for instance. If we solely go for innovation and ignore their needs, many who can’t adjust to change will suffer. Technology’s main objective is to innovate, however, I believe it also exists to help such people. Looking upon, there is not much done to execute this technology to aid the blind and visually impaired. With this thought in mind, I came up with an idea of an app that would provide assistance to those who are blind or visually impaired, in recognising object and colour in real-time.
This idea is quite personal to me, as one of my close family relatives has started losing vision. Seeing them struggle to work out even the basic tasks triggered me to create something that would help the blind community regain their sense of independence. I believe that implementing the idea of artificial intelligence to aid the blind community, will help them do so.
About the App
My application would be called “Ikshana”, literally meaning sight in Sanskrit. Other meanings that could be interpreted from this word are eye, viewing, caring for and superintendence. I intend to provide features such as text recognition, image recognition (includes identification of different types of food, plants, flowers and animal species), recognition of packaged products and money recognition (includes Indian Rupees and US Dollars). In addition to that, it will also give details of the nearby hospitals and be able to book appointments as well. It will use the phone’s camera and voiceover functions to identify the object or anything by saying it out loud. The app would be available on both Android and iOS operating systems.
There are a few applications available on the Play Store and App store for the visually impaired and blind which provide similar features as Ikshana. Aipoly and TapTapSee are some of them. Both serve object identification feature and give out output in the form of voice. Ikshana would not only identify most type of objects but also provide the feature of information on hospitals booking appointments. This becomes quite helpful to the person as they won’t be needing any further assistance or depend on anyone to book a hospital appointment. It can simply be done by providing their voiceover without actually physically going to the hospital.
It is quite obvious that this app will majorly cater to people with visual disabilities, hence the target customers would be restricted to the blind and visually impaired community. India is home to about 12 million blind people. However, only about a fraction of them would be able to afford a smartphone or work with technology, which reduces the market size. Even so, I aim to reach out to the 40 million suffering from blindness across the globe. The app will be available in six languages, that are, English, Hindi, Chinese, French, Arabic and German. As the app expands, more languages can be added in the future.
Requirements
Functional:
- The system must identify the objects placed in front of it and the name of the object should be given as a voice output. It must detect the colour and give out brief information about the object.
- The input must only be taken as the voice of the user.
- The output should in the form of a voice output as well as in a text format.
Non-functional:
- Information of various hospitals should be provided by the system.
- Learning libraries such as Tensorflow, OpenCV and Microsoft Cognitive Services will be used in the development of the system.
- The programming languages picked for this process are Java and Python.
- Image recognition should have an accuracy of at least 92%.
Implementation
- Conversion of speech to text. Translation of speech inputs to text outputs will be done through a statistical modelling framework. For this, Hidden Markov Model is being introduced specifically.
- The translated text is then converted to a collection of tokens using a tokenizer.
- These collections of tokens are evaluated by a syntactic analyser to provide a syntactically accurate sentence.
- Suitable actions are taken when user commands are detected using these parsed tokens and doing question and intent analysis.
- Object detection using the camera of the device will be done with the help of Microsoft Cognitive Services libraries and finally, the results are analysed.
- The system converts the text output to synthesised speech using speech synthesis.
Defining capital
I plan to develop this app with the help of at most three professional app developers to improve the quality of the app service and fasten its process. While there are not any equipment or hardware required for this project, the only costs that would be incurred are the salaries of the app developers and advertisement costs. The total budget of this project would come around INR 1.8 lakhs to 2.0 lakhs, which includes four months of salaries for the app developers and the rest would be used up for advertising and marketing.
Subscription plans and pricing
The app will be available on both Google Play Store and App Store for free to download for everyone. However, there will be in-app purchases available in the app. For the free version of the app, common object recognition and colour detection will be available. To utilise all the features of the app, a subscription plan is decided that the users can purchase. Users can purchase either a monthly subscription or a yearly one. The monthly subscription will start with a 3 weeks free trial, after which a fee of ₹ 299 per month will be charged. However, if the users wish to pay only once, a yearly subscription is also available which they can buy for ₹ 2990.
Marketing
I plan to use both above the line and below the line marketing strategies to reach out to people. I will advertise the application using strategies that are both targeted and non-targeted to spread awareness about it to a wider audience. Since the targeted audience is mainly blind and visually disabled, the advertising activities are limited. I plan to approach the audience by the means of radio, social media and if possible, spreading awareness about the app service in certain hospitals. I chose to advertise the service in hospitals through the doctors or add banners in the building. This is where most of the targeted customers can be found. People can get to know about this service through doctors. This method is quite trustable and will certainly help the ones who wish to carry out their daily tasks on their own. Next, I plan to put an advertisement on the radio as this is something the blind can access to easily. By putting an ad on one of the radio stations, the people can know about the app. The other means of advertising I will be using is social media. By setting up accounts of the application service on various social media platforms, a wider reach can be achieved. It can help to reach out to everyone whether they require the help of this service or not. However, it can still be helpful as more will get to know about it and recommend it to those who need this type of assistance in their lives.
Risk Analysis
There is always some sort of limitations and errors that can happen while working on a project. In the case of Ikshana, there are multiple things that can go wrong while developing it. For instance, the app may not be able to provide a strong user experience for some. Certain visuals or animations may seem fine to us but it might be too strong or weak for those who will actually use the app. However, this might be different for everyone.
Another thing that can go wrong is not doing enough testing. With a timebound, it can happen so that enough tests are not conducted. This may lead to bugs or errors that would cause problems when the app is launched. This can still be solved later on, but it can have an effect on the trust of the customers.
As mentioned before, the app can only identify objects at an accuracy of 92%, which means it won’t be able to give out fully accurate results. While detecting an object, certain factors like angles or the distance from the object can cause confusion while generating results. This can be a limitation as the accuracy cannot be increased. If it is increased, the app would use up too much battery and for some devices, it might not work at all. To make it work efficiently and work on simple devices, the accuracy is kept slightly low so that it provides fairly good results keeping the efficiency in mind.s
Action Plan
Drafting of the idea, concept- by November
Recruiting engineers for app development- by November
Defining overall costs of the project- by December
Drafting of features and core functions- by December
Defining programming language- by December
Defining platforms- by December
Research on alternatives or competitors- by December
Creating a mock-up- by January
Define UI/UX- by January
Fix UI/UX- by January
Testing and deployment- by February
Defining subscription plans and pricing of the features- by February
Launch- by February
