Digital aids based on image or object classification are very useful, providing visually impaired people help recognize nearby objects. Creating a system that provides audio feedback of the surrounding objects to the visually impaired, for the advancements in computer vision computing technology. In order to help the visually impaired and the blind, this study investigates Image Classification methods, as it can be very helpful for the their safety, quality of life and constant independence from other people.The Image Classification module recognizes 30 objects such as apple, bed, banana, cars, chair, dog, human, door_closed, door_opened, human_with_mask, tree, zebra_crossing, etc. and generates an audio feedback to the user by converting the classified text to speech using python. Android Studio development environment has been used for Android development. Work on the layout of the App has also been done using Android Studio.Online available dataset has been used for the evaluation of the Image Classification model. To increase the accuracy, clear, precise and more number of images in the data set have been used. To teach the model to classify images using files or webcam of Teachable Machine [5] is used. The models built with Teachable Machine are authentic TensorFlow.js models which function anywhere, where javascript is present, which makes them compatible with Glitch, P5.js, Node.js, and other tools. When we create our own classes, they transform into the final layer or segment of a neural network that has already been trained. Particularly, the image models are picking up knowledge from pre-trained mobilenet models.