From c2f7b5bad38f45baff40256db6c7d526b143d149 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Kaan=20=C3=87olak?= <kaancolak95@gmail.com>
Date: Mon, 13 Nov 2023 22:52:24 +0300
Subject: [PATCH] docs(how-to-guides): add training docs for centerpoint
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Kaan Çolak <kaancolak95@gmail.com>
---
 .../training-models.md                            | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/docs/how-to-guides/training-machine-learning-models/training-models.md b/docs/how-to-guides/training-machine-learning-models/training-models.md
index 51c9091918d..4b9d3e96747 100644
--- a/docs/how-to-guides/training-machine-learning-models/training-models.md
+++ b/docs/how-to-guides/training-machine-learning-models/training-models.md
@@ -28,10 +28,17 @@ an example dataset containing three distinct classes (green, yellow, red), which
 
 Detailed instructions for training the traffic light classifier model can be found **[here](https://github.com/autowarefoundation/autoware.universe/blob/main/perception/traffic_light_classifier/README.md)**.
 
-<!--
+## Training CenterPoint 3D object detection model
 
-Training traffic light detection model and lidar CenterPoint model will be added there.
+The CenterPoint 3D object detection model within the Autoware has been trained using the **[open-mmlab/mmdetection3d](https://github.com/open-mmlab/mmdetection3d)** repository. CenterPoint
+implementation of mmdetection3d uses 10 input feature for PointPillars voxel encoder. However, Autoware employs 9 pillar features for CenterPoint, in accordance with the original research paper. Therefore, we have
+fork the original repository and made the necessary code modifications to support the use of 9 PointPillar input features. The forked repository can be found **[here](https://github.com/autowarefoundation/mmdetection3d)**.
 
--->
+To train custom CenterPoint models and convert them into ONNX format for deployment in Autoware, please refer to the instructions provided in the README file included with the
+**"lidar_centerpoint"** package. These instructions will provide a step-by-step guide for training CenterPoint model.
 
-## Training CenterPoint 3D object detection model
+In order to assist you with your training process, we have also included an example dataset in the Tier4 dataset format.
+This dataset contains 600 lidar frames and covers 5 classes, including 6905 cars, 3951 pedestrians, 75 cyclists, 162 buses, and 326 trucks.
+You can utilize this example dataset to facilitate your training efforts.
+
+Detailed instructions for training the CenterPoint model can be found **[here](https://github.com/autowarefoundation/autoware.universe/blob/main/perception/lidar_centerpoint/README.md)**.