I implemented this project in the hope so as to understand what a neural network is in basic idea. This gave me the pleasure of implementing my first AI project.
Image recognition using deep learning is a long and complex process to be coded from scratch. But it has a great welcome from coders of varied sectors. So, it is made easy by making them open source such that it can be used in anyone's code. Resnet50 is one such model that i have used to perform image recognition.
The program listens to your command. When the user says "Capture", the camera starts functioning and captures a picture.
Then, the captured image is given as an input for the resnet50 model. There, the objects in the image are recognised.
This recognised objects are then localised and thus found in which part of the image it is located. The output is in speech format that says which objects are in left, center and right.
This can be integrated alongside a map service such as the google maps. In that case, not just about objects ahead, the program can also give voice commands on guiding to our location. This could be of great use to blind people. It can act as a virtual eye.