ROS package for the Perception (Sensor Processing, Detection, Tracking and Evaluation) of the KITTI Vision Benchmark
Sticking to this folder structure is highly recommended:
~ # Home directory
├── catkin_ws # Catkin workspace
│ ├── src # Source folder
│ └── SARosPerceptionKitti # Repo
├── kitti_data # Dataset
│ ├── 0060 # Demo scenario 0060
│ │ ├── segmented_semantic_images # Folder for preprocessed semantic segmented images
│ │ └── synchronized_data.bag # Synchronized ROSbag file
- Install ROS and create a catkin workspace in your home directory:
mkdir -p ~/catkin_ws/src
- Clone this repository into the catkin workspace's source folder (src) and build it:
cd ~/catkin_ws/src
git clone https://github.com/appinho/SARosPerceptionKitti.git
cd ~/catkin_ws
catkin_make
source devel/setup.bash
- Download a preprocessed scenario and unzip it into a separate
kitti_datadirectory, also stored under your home directory:
mkdir ~/kitti_data && cd ~/kitti_data/
mv ~/Downloads/0060.zip .
unzip 0060.zip
rm 0060.zip
- Launch one of the following ROS nodes to perform and visualize the pipeline (Sensor Processing -> Object Detection -> Object Tracking) step-by-step:
source devel/setup.bash
roslaunch sensor_processing sensor_processing.launch home_dir:=/home/YOUR_USERNAME
roslaunch detection detection.launch home_dir:=/home/YOUR_USERNAME
roslaunch tracking tracking.launch home_dir:=/home/YOUR_USERNAME
- Default parameters:
- scenario:=0060
- speed:=0.2
- delay:=3
Without assigning any of the abovementioned parameters the demo scenario 0060 is replayed at 20% of its speed with a 3 second delay so RViz has enough time to boot up.
- Write the results to file and evaluate them:
roslaunch evaluation evaluation.launch home_dir:=/home/YOUR_USERNAME
cd ~/catkin_ws/src/SARosPerceptionKitti/benchmark/python
python evaluate_tracking.py
| Class | MOTA | MOTP | MOTAL | MODA | MODP |
|---|---|---|---|---|---|
| Car | 0.881119 | 0.633595 | 0.881119 | 0.881119 | 0.642273 |
| Pedestrian | 0.546875 | 0.677919 | 0.546875 | 0.546875 | 0.836921 |
If you have any questions, things you would love to add or ideas how to actualize the points in the Area of Improvements, send me an email at [email protected] ! More than interested to collaborate and hear any kind of feedback.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent