Deep learning based application for image correction

I’m using the Jetson Nano board with Kernel : 4.9.140

The accelerometer supported is LSM6DSM, trying further to add any application that can sense the motion of camera from Accelerometer values, and correct the image if its blurred or having larger shakes. Found a similar application in android at : http://jasonmcreynolds.com/?p=388, trying same in Ubuntu systems.

I’ve reckoned that Deep Learning is supported by TensorFlow+CUDA. Thus, any help on the queries above is welcomed, thanks.

I referred the LSM6DSM support in mode4, where the IMU can be connected to Camera Sensor and similar functionalities can be achieved.
Any input on how the DeepStream support on Jetson Nano can help in image correction.