Briefly introduce my CS(Computer Science) final project. In recent years, AI(Artificial Intelligence) products and relative theorems have sprung up, which motivated us to apply a certain branch of AI theorems called "Deep Learning" as a method for implementing our final project. Our goal is to make an image caption generator for those who have visual disability . The project is divided two parts. The first part will combine CNN and LSTM model to train our machine so that it has the ability to describe after receiving images. In the second part, we intend to use raspberry pi as a device which will take a picture when something trigger ultrasonic sensor. The device will return the picture taken by camera to the trained machine and receive the result analyzed by it in the form of MP3 file. Finally, the blindman can roughly know the sight in front of him through Bluetooth headset.