Carla Simulator YOLOV3 Object Detection Sample

Simple sample for Carla Simulator, object detection with bounding box application with yoloV3 (tensorflow 1.15.x)

YoloV3 Tensorflow implementation forked from: https://github.com/YunYang1994/tensorflow-yolov3 (branch: dependabot/pip/docs/tensorflow-gpu-1.15.2)

My Environment

Windows 10 Carla 0.9.9 python 3.7.5 tensorflow-gpu 1.15.0 pygame 1.9.6 opencv-python 4.2.0.34 numpy 1.18.3 pillow 7.1.2

Project Directory Structure

.CARLA_0.9.9            
├── WindowsNoEditor
│   │   ├── CarlaUE4
│   │   ├── Co-Simulation
│   │   ├── Engine
│   │   ├── HDMaps
│   │   ├── PythonAPI
│   │   │   ├── carla
│   │   │   ├── util
│   │   │   ├── examples
│   │   │   │ 	├── yolov3_object_detection.py
│   │   │   │ 	├── tensorflow_yolov3    
│   │   │   │ 	│   │  	├── carla
│   │   │   │ 	│   │	│   ├── utils.py            

Setup

1. Open a command line Go to the Carla Simulator examples path ..\CARLA_0.9.9\WindowsNoEditor\PythonAPI\examples

   Clone this repo without project folder with the below section

   git init
   git remote add origin https://github.com/umtclskn/Carla_Simulator_YOLOV3_Object_Detection.git
   git pull origin master
   git submodule update --init --recursive
   

2. Download COCO weights from this link:

   https://github.com/YunYang1994/tensorflow-yolov3/releases/download/v1.0/yolov3_coco.tar.gz
   

   extract this file under the below path:

   ..\CARLA_0.9.9\WindowsNoEditor\PythonAPI\examples\tensorflow-yolov3\checkpoint

---

3. (type these command at the ..\CARLA_0.9.9\WindowsNoEditor\PythonAPI\examples\tensorflow-yolov3)

   cd..
   python convert_weight.py
   python freeze_graph.py
   

   see the

4. Open CarlaEU4.exe (..\CARLA_0.9.9\WindowsNoEditor)

5. Run spawn actor python file for adding pedestrians or vehicles.

   python spawn_npc.py

6. Start detecting vehicles, pedestrians or bicycles.

   python yolov3_object_detection.py