Update VisionNode_CodeSummary.md

doc/research/paf24/perception/VisionNode_CodeSummary.md

@@ -3,6 +3,7 @@
 The `VisionNode` class is designed to perform object detection and segmentation tasks using both PyTorch and Ultralytics models. It is structured to publish detection and segmentation results in ROS.
 
 ## Table of Contents
+
 - [Table of Contents](#table-of-contents)
 - [Overview](#overview)
 - [Class Initialization](#class-initialization)
@@ -24,9 +25,11 @@ The `VisionNode` class is designed to perform object detection and segmentation
 
 
 ## Overview
+
 The `VisionNode` class leverages both PyTorch and Ultralytics pretrained models for object detection and segmentation, integrating features such as bounding boxes, segmentation masks, and distance calculations based on LIDAR data.
 
 ## Class Initialization
+
 Upon instantiation, `VisionNode` initializes various configurations, such as:
 - **Model Dictionary** (`self.model_dict`): Holds model configurations with different detection and segmentation models from PyTorch and Ultralytics.
 - **Camera Configurations**: Subscribes to topics for front, rear, left, and right camera views if specified.
@@ -37,44 +40,55 @@ Upon instantiation, `VisionNode` initializes various configurations, such as:
 ## Setup Functions
 
 ### Camera and Distance Array Subscriptions
+
 These functions subscribe to the camera topics, allowing the node to receive images from multiple camera angles and perform image processing tasks.
 
 - **`setup_camera_subscriptions`**: Subscribes to specified camera angles.
 - **`setup_dist_array_subscription`**: Subscribes to LIDAR depth data.
 
 ### Camera Publishers
+
 **`setup_camera_publishers`** sets up publishers for each camera angle. Each angle has a unique topic to publish segmented images back to ROS.
 
 ### Object Distance Publishers
+
 **`setup_object_distance_publishers`** creates a publisher that outputs a list of detected objects and their distances.
 
 ### Traffic Light Publishers
+
 **`setup_traffic_light_publishers`** sets up a publisher for detected traffic light images.
 
 ## Main Processing Functions
 
 ### Handling Camera Images
+
 **`handle_camera_image`** receives an image message, selects the appropriate prediction framework (PyTorch or Ultralytics), and processes the image. After processing, it publishes the result on the appropriate topic based on the camera angle.
 
 ### Distance Array Handling
+
 **`handle_dist_array`** receives LIDAR data and updates the `self.dist_arrays` attribute with the latest depth image data.
 
 ### Prediction with PyTorch
+
 **`predict_torch`** preprocesses images, applies the PyTorch model, and draws either bounding boxes or segmentation masks based on the model type (detection or segmentation).
 
 ### Prediction with Ultralytics
+
 **`predict_ultralytics`** preprocesses images and applies an Ultralytics model, handling object detection and drawing bounding boxes on the image. This function includes distance calculation logic using LIDAR data within the bounding boxes.
 
 ### Traffic Light Processing
+
 **`process_traffic_lights`** identifies traffic lights based on object detection outputs. It crops the detected regions and publishes segmented traffic light images aswell as distance information to a separate topic.
 
 ### Bounding Box and Segmentation Mask Creation
+
 - **`create_mask`**: Generates a segmentation mask for PyTorch segmentation models.
 - **`apply_bounding_boxes`**: Draws bounding boxes on detection models and labels each bounding box with its respective object type and distance information.
 
 ## Utility Functions
 
 ### Minimum X and Y Calculations
+
 - **`min_x`**: Calculates the minimum x-distance for objects in LIDAR data, representing the closest object in the forward direction.
 - **`min_abs_y`**: Calculates the minimum y-distance  in absolute terms, representing the closest object sideways.