Dhruv Image Recognizer: Honest Review

dhruvtalksstech
Nov 5, 2022
4 min read

Updated: Dec 3, 2022

(https://play.google.com/store/apps/details?id=com.purpletutor.info.dhruvimagerecognizerapp) - Link to download the app

It is an app dedicated to identifying the objects that we take pictures of. It uses AI and the python codes written by me to run the picture against webpages and neural and AI engines to find similar results and then it tries to match the image with a string of pictures of objects that it thinks are similar. I created the app using Thunkable which is an online app-building website. I kept the interface relatively simple and it just requires clicking on the “capture photos” button; then, it asks for permission to access the camera and storage. Privacy was the prime concern for me and that is why I didn’t code hidden trackers and bots which track user data to sell it later (trackers and cross-trackers are embedded in all apps and this is a very common practice in the industry.

It may be in the form of cookies as well. Image Recognizer never stores such data. Although it seeks permission to access the media on your phone, the app never goes through any of your photos and files to protect your privacy. It is strictly against my policy. Yes, the app is not as polished and accurate as something like Google Lens but is fairly precise and I hope to improve it with future updates. Before writing the codes itself for the app, I designed the user interface as background, colors, margins, interface, buttons, effect after pressing the button, font, etc. The codes utilized “if, when, then, call, then do” conditions and statements.

It also uses Microsoft’s AI and search engine. Before releasing the app, I also tested it on about 100 objects to check if it was working fine. For the objects that it was recognized incorrectly, I had to manually override the code for those objects specifically to run it against more strings when it detected a similar object in the public version of the app. For the image captured to be recognized as an object, the picture had to reach a certain threshold of similarity in terms of the strings and range of similar objects it corresponded to. I then measured the objects captured in terms of confidence level and error rate.

Even after identifying the object, it has to be described correctly to the user. So, for the actual description that is displayed, I altered the description not to just identify the object but be more natural and accommodative of the surroundings. For eg- an image of an AC remote would describe as “a person holding a remote control” if it detects skin around the object. I hope to continue testing the app and roll out updates in the future to improve object recognition with special emphasis on accurate description.

Yes, the Image recognizer app is very much like Google Lens but it differs in the fundamentals in that it can also give an accurate description of the object shown rather than pulling it on the web and showing buying options. My app is also not as smooth right now as Google Lens but it is the best I could do given the resources and level of expertise at my disposal. I enjoy using it very much and consider it a baby step into the world of coding and hardcore Computer Science. I have always been very interested in tech and always wanted to do something outside the world of personal consumer electronics.

I find it fascinating that a few lines of code have the potential to change the world. I plan to gain more knowledge of the field and create more such apps and maybe websites in the future. I can't wait to share more of my work with you and it brings me immense pleasure as I write this blog to share my first ever real creation and as I like to say "work of art". I also wish to address a possible flag. The app is only available for Google Play Store currently and not Apple's App Store as of now as the process to have an App Store is a very tedious one and requires expertise to get approval and fill out the details and a lot of agreements. It is also a lot stricter in its requirements and demands a lot of signings from the app developer.

The app is also available on the Amazon App Store - https://www.amazon.com/gp/product/B0B9T384WV

The app has also been featured in 3 newspapers to date - Dainik Jagran, Dainik Hint, and Yug Karvat. Privacy was at the very center of this app since privacy and security are very important to me. The app does not collect any data from the phone and does not store the images using the app. I have a strict no-logs policy and the app does not ask for any unrequired permissions. The app has 70 reviews and over 1200 downloads as of reviews as of October 29, 2022, and an overall app rating of 4.9 stars.