feat(trajectory-rendering): add README to describe the visualization functions

Signed-off-by: Arseni10Lk <arseniy230606@gmail.com>

Author: Arseni10Lk

Model/visualization/README.md

@@ -0,0 +1,55 @@
+# Trajectory Visualization #
+
+This is implemented as a class. It contains 4 functions:
+* `accel_and_curv_to_meters_trajectory`
+* `meters_to_pixels_trajectory`
+* `overlay_the_trajectory_with_map`
+* `render_trajectory_map_tile`
+
+## Complete function ##
+
+### `render_trajectory_map_tile`
+
+Integrates predicted trajectory into metric coordinates and 
+draws them onto the raw BEV map tile.
+
+It takes four inputs:
+* `action_sequence`: (128, ) flattened (64, 2) $[acceleration, curvature]$ tensor.
+It is the exact format of the trajectory outputted by `model()` function.
+* `current_speed`: Scalar float from the egomotion history. 
+This should be extracted from the egomotion history
+* `map_image`: A map tile, not normalized. 
+Ideally, it should follow L2D format. 
+If the dataset does not provide maps directly, but provides GPS history, 
+please use already existing [map generation function](https://github.com/autowarefoundation/auto_e2e/tree/main/Model/data_parsing/map_rendering)
+to get a map tile.
+* `radius_m`: The metric boundary of the `map_image` in meters.
+
+Returns:
+* A new PIL Image with the trajectory drawn on it.
+
+## Helper functions ##
+
+### `accel_and_curv_to_meters_trajectory`
+Takes an action sequence containing pairs of acceleration ($m/s^2$) and curvature ($rad/m$), the ego vehicle's current speed, and a defined number of future timesteps. 
+It integrates these values over time using a kinematic model to produce a tensor of metric $(X, Y)$ coordinates. The ego vehicle is positioned at the origin $(0.0, 0.0)$, with forward movement mapped to the positive $Y$-axis.
+
+### `meters_to_pixels_trajectory`
+Converts the metric trajectory tensor into 2D image pixel coordinates $(U, V)$. It scales the coordinates based on the pixel dimensions of the provided map image and the metric boundary `radius_m`.
+
+### `overlay_the_trajectory_with_map`
+Uses PIL's `ImageDraw` module to render the trajectory as a continuous green line directly onto a copy of the BEV map image. It also draws a red circle at the trajectory's origin to indicate the ego vehicle's current position.
+
+## Dependencies ##
+
+Core visualization logic only requires standard machine learning and image processing libraries:
+
+```text
+Pillow>=5.3.0
+```
+
+If you wish to run the live visualization script (`--live`) to test predictions using real dataset records, you will additionally need the [L2D dependencies](https://github.com/autowarefoundation/auto_e2e/tree/main/Model/data_parsing/l2d) 
+
+## TODO ##
+
+A function to draw the trajectory on the camera view is yet to be added.