Reverse time migration

This example locates at examples/rtm/marmousi, we use a towed acqusition system with marmousi model to perform rtm.

Generate geometry and observed data

Change to the example folder:
```
cd examples/rtm/marmousi
```
Generate the geometry and observed data:
```
python generate_model_geometry.py
```
Perform forward modeling:
```
sh forward.sh
```
Remove first arrivals

Generate the data mask to remove the first arrivals.
```
python generate_datamask.py
```
Perform RTM

Set the loss in .sh file to rtm, and set epoch in .yml file to 1 to perform one fwi iteration for calculating the gradient.

Run rtm with true model.
```
sh rtm_truemodel.sh
```
Run rtm with initial model.
```
sh rtm_initmodel.sh
```
Get the migration results Show the gradietns and migration results.
```
python show results.py
```
The automatic differention calculated gradients by backpropagating the observed data are shown as follows:

The laplace filtered migration sections are shown below: