Author: Luigi Freda
- pySLAM v2.2
pySLAM is a python implementation of a Visual Odometry (VO) pipeline for monocular, stereo and RGBD cameras. It supports many classical and modern local features, and it offers a convenient interface for them. Moreover, it collects other common and useful VO and SLAM tools.
I released the first version of pySLAM (v1) for educational purposes, for a computer vision class I taught. I started developing it for fun as a python programming exercise, during my free time, taking inspiration from some repos available on the web.
Main Scripts:
-
main_vo.py
combines the simplest VO ingredients without performing any image point triangulation or windowed bundle adjustment. At each step$k$ ,main_vo.py
estimates the current camera pose$C_k$ with respect to the previous one$C_{k-1}$ . The inter-frame pose estimation returns$[R_{k-1,k},t_{k-1,k}]$ with$\Vert t_{k-1,k} \Vert=1$ . With this very basic approach, you need to use a ground truth in order to recover a correct inter-frame scale$s$ and estimate a valid trajectory by composing$C_k = C_{k-1} [R_{k-1,k}, s t_{k-1,k}]$ . This script is a first start to understand the basics of inter-frame feature tracking and camera pose estimation. -
main_slam.py
adds feature tracking along multiple frames, point triangulation, keyframe management and bundle adjustment in order to estimate the camera trajectory up-to-scale and build a map. It's still a VO pipeline but it shows some basic blocks which are necessary to develop a real visual SLAM pipeline. -
main_feature_matching.py
shows how to use the basic feature tracker capabilities (feature detector + feature descriptor + feature matcher) and allows to test the different available local features. Further details here. -
main_map_viewer.py
allows to reload a saved map and visualize it.
You can use this framework as a baseline to play with local features, VO techniques and create your own (proof of concept) VO/SLAM pipeline in python. When you test it, consider that's a work in progress, a development framework written in Python, without any pretence of having state-of-the-art localization accuracy or real-time performances.
Enjoy it!
First, clone this repo and its modules by running
$ git clone --recursive https://github.com/luigifreda/pyslam.git
$ cd pyslam
Then, use the available specific install procedure according to your OS. The provided scripts will create a single python environment that is able to host all the supported local features!
- Python 3.8.10
- OpenCV >=4.8.1 (see below)
- PyTorch 2.3.1
- Tensorflow 2.13.1
- Kornia 0.7.3
- Rerun
If you run into troubles or performance issues, check this TROUBLESHOOTING file.
Follow the instructions reported here for creating a new virtual environment pyslam
with venv. The procedure has been tested on Ubuntu 18.04, 20.04, 22.04 and 24.04.
If you prefer conda, run the scripts described in this other file.
Follow the instructions in this file. The reported procedure was tested under Sonoma 14.5 and Xcode 15.4.
If you prefer docker or you have an OS that is not supported yet, you can use rosdocker:
- with its custom
pyslam
/pyslam_cuda
docker files and follow the instructions here. - with one of the suggested docker images (ubuntu*_cuda or ubuntu*), where you can build and run pyslam.
The provided install scripts take care of installing a recent opencv version (>=4.8) with its non-free modules enabled (see for instance install_pip3_packages.sh, which is used with venv under Ubuntu, or the install_opencv_python.sh under mac).
How to check your installed OpenCV version:
$ python3 -c "import cv2; print(cv2.__version__)"
How to check if you have non-free OpenCV module support (no errors imply success):
$ python3 -c "import cv2; detector = cv2.xfeatures2d.SURF_create()"
If you run into issues or errors during the installation process or at run-time, please, check the TROUBLESHOOTING.md file.
Once you have run the script install_all_venv.sh
(follow the instructions above according to your OS), you can open a new terminal and run:
$ . pyenv-activate.sh # Activate pyslam python virtual environment. This is just needed once in a new terminal.
$ ./main_vo.py
This will process a default KITTI video (available in the folder videos
) by using its corresponding camera calibration file (available in the folder settings
), and its groundtruth (available in the same videos
folder). You can stop main_vo.py
by focusing on the Trajectory window and pressing the key 'Q'.
Note: As explained above, the basic script main_vo.py
strictly requires a ground truth.
In order to process a different dataset, you need to set the file config.yaml
:
- Select your dataset
type
in the sectionDATASET
(further details in the section Datasets below for further details). This identifies a corresponding dataset section (e.g.KITTI_DATASET
,TUM_DATASET
, etc). - Select the
sensor_type
(mono
,stereo
,rgbd
) in the chosen dataset section. - Select the camera
settings
file in the dataset section (further details in the section Camera Settings below). - The
groudtruth_file
accordingly (further details in the section Datasets below and check the filesground_truth.py
andconvert_groundtruth.py
).
Similarly, you can test main_slam.py
by running:
$ . pyenv-activate.sh # Activate pyslam python virtual environment. This is just needed once in a new terminal.
$ ./main_slam.py
This will process a default KITTI video (available in the folder videos
) by using its corresponding camera calibration file (available in the folder settings
). You can stop it by focusing on the opened Figure 1 window and pressing the key 'Q'.
Note: Due to information loss in video compression, main_slam.py
tracking may peform worse with the available KITTI videos than with the original KITTI image sequences. The available videos are intended to be used for a first quick test. Please, download and use the original KITTI image sequences as explained below.
If you just want to test the basic feature tracker capabilities (feature detector + feature descriptor + feature matcher) and get a tast of the different available local features, run
$ . pyenv-activate.sh # Activate pyslam python virtual environment. This is just needed once in a new terminal.
$ ./main_feature_matching.py
In any of the above scripts, you can choose any detector/descriptor among ORB, SIFT, SURF, BRISK, AKAZE, SuperPoint, etc. (see the section Supported Local Features below for further information).
Some basic test/example files are available in the subfolder test
. In particular, as for feature detection/description, you may want to take a look at test/cv/test_feature_manager.py too.
When you run the script main_slam.py
:
- The current map can be saved into the file
map.json
by pressing the buttonSave
on the GUI. - The saved map can be reloaded and visualized into the GUI by running:
$ . pyenv-activate.sh # Activate pyslam python virtual environment. This is just needed once in a new terminal.
$ ./main_map_viewer.py
Relocalization in a loaded map is a WIP.
Estimated trajectories can be saved in three different formats: TUM (The Open Mapping format), KITTI (KITTI Odometry format), and EuRoC (EuRoC MAV format). To enable trajectory saving, open config.yaml
and search for the SAVE_TRAJECTORY
: set save_trajectory: True
, select your format_type
(tum
, kitti
, euroc
), and the output filename. For instance for a tum
format output:
SAVE_TRAJECTORY:
save_trajectory: True
format_type: tum
filename: kitti00_trajectory.txt
Some quick information about the non-trivial GUI buttons of main_slam.py
:
Step
: Enter in the Step by step mode. Press the buttonStep
a first time to pause. Then, press it again to make the pipeline process a single new frame.Save
: Save the map into the filemap.json
. You can visualize it back by using the script/main_map_viewer.py
(as explained above).Draw GT
: In the case a groundtruth is loaded (e.g. with KITTI, TUM, EUROC datasets), you can visualize it by pressing this button. The groundtruth trajectory will be visualized and progressively aligned to the estimated trajectory.
At present time, the following feature detectors are supported:
- FAST
- Good features to track
- ORB
- ORB2 (improvements of ORB-SLAM2 to ORB detector)
- SIFT
- SURF
- KAZE
- AKAZE
- BRISK
- AGAST
- MSER
- StarDector/CenSurE
- Harris-Laplace
- SuperPoint
- D2-Net
- DELF
- Contextdesc
- LFNet
- R2D2
- Key.Net
- DISK
- ALIKED
- Xfeat
- KeyNetAffNetHardNet (KeyNet detector + AffNet + HardNet descriptor).
The following feature descriptors are supported:
- ORB
- SIFT
- ROOT SIFT
- SURF
- AKAZE
- BRISK
- FREAK
- SuperPoint
- Tfeat
- BOOST_DESC
- DAISY
- LATCH
- LUCID
- VGG
- Hardnet
- GeoDesc
- SOSNet
- L2Net
- Log-polar descriptor
- D2-Net
- DELF
- Contextdesc
- LFNet
- R2D2
- BEBLID
- DISK
- ALIKED
- Xfeat
- KeyNetAffNetHardNet (KeyNet detector + AffNet + HardNet descriptor).
You can find further information in the file feature_types.py. Some of the local features consist of a joint detector-descriptor. You can start playing with the supported local features by taking a look at test/cv/test_feature_manager.py
and main_feature_matching.py
.
In both the scripts main_vo.py
and main_slam.py
, you can create your favourite detector-descritor configuration and feed it to the function feature_tracker_factory()
. Some ready-to-use configurations are already available in the file feature_tracker.configs.py
The function feature_tracker_factory()
can be found in the file feature_tracker.py
. Take a look at the file feature_manager.py
for further details.
N.B.: you just need a single python environment to be able to work with all the supported local features!
You can use 5 different types of datasets:
Dataset | type in config.yaml |
---|---|
KITTI odometry data set (grayscale, 22 GB) | type: KITTI_DATASET |
TUM dataset | type: TUM_DATASET |
EUROC dataset | type: EUROC_DATASET |
Video file | type: VIDEO_DATASET |
Folder of images | type: FOLDER_DATASET |
pySLAM code expects the following structure in the specified KITTI path folder (specified in the section KITTI_DATASET
of the file config.yaml
). :
├── sequences
├── 00
...
├── 21
├── poses
├── 00.txt
...
├── 10.txt
-
Download the dataset (grayscale images) from http://www.cvlibs.net/datasets/kitti/eval_odometry.php and prepare the KITTI folder as specified above
-
Select the corresponding calibration settings file (parameter
[KITTI_DATASET][cam_settings]
in the fileconfig.yaml
)
pySLAM code expects a file associations.txt
in each TUM dataset folder (specified in the section TUM_DATASET:
of the file config.yaml
).
-
Download a sequence from http://vision.in.tum.de/data/datasets/rgbd-dataset/download and uncompress it.
-
Associate RGB images and depth images using the python script associate.py. You can generate your
associations.txt
file by executing:
$ python associate.py PATH_TO_SEQUENCE/rgb.txt PATH_TO_SEQUENCE/depth.txt > associations.txt
- Select the corresponding calibration settings file (parameter
TUM_DATASET: cam_settings:
in the fileconfig.yaml
)
-
Download a sequence (ASL format) from http://projects.asl.ethz.ch/datasets/doku.php?id=kmavvisualinertialdatasets (check this direct link)
-
Select the corresponding calibration settings file (parameter
EUROC_DATASET: cam_settings:
in the fileconfig.yaml
)
The folder settings
contains the camera settings files which can be used for testing the code. These are the same used in the framework ORB-SLAM2. You can easily modify one of those files for creating your own new calibration file (for your new datasets).
In order to calibrate your camera, you can use the scripts in the folder calibration
. In particular:
- use the script
grab_chessboard_images.py
to collect a sequence of images where the chessboard can be detected (set the chessboard size therein, you can use the calibration patterncalib_pattern.pdf
in the same folder) - use the script
calibrate.py
to process the collected images and compute the calibration parameters (set the chessboard size therein)
For further information about the calibration process, you may want to have a look here.
If you want to use your camera, you have to:
- calibrate it and configure WEBCAM.yaml accordingly
- record a video (for instance, by using
save_video.py
in the foldercalibration
) - configure the
VIDEO_DATASET
section ofconfig.yaml
in order to point to your recorded video.
For a comparison of the trajectories estimated by pySLAM and by ORB-SLAM3, see this trajectory comparison notebook.
Note that pySLAM pose estimates are saved online: At each frame, the current pose estimate is saved. On the other end, ORB-SLAM3 pose estimates are saved at the end of the dataset playback: That means each pose estimate
If you like pySLAM and would like to contribute to the code base, you can report bugs, leave comments and proposing new features through issues and pull requests on github. Feel free to get in touch at luigifreda(at)gmail[dot]com. Thank you!
Suggested books:
- Multiple View Geometry in Computer Vision by Richard Hartley and Andrew Zisserman
- An Invitation to 3-D Vision by Yi-Ma, Stefano Soatto, Jana Kosecka, S. Shankar Sastry
- Computer Vision: Algorithms and Applications, by Richard Szeliski
- Deep Learning, by Ian Goodfellow, Yoshua Bengio and Aaron Courville
- Neural Networks and Deep Learning, By Michael Nielsen
Suggested material:
- Vision Algorithms for Mobile Robotics by Davide Scaramuzza
- CS 682 Computer Vision by Jana Kosecka
- ORB-SLAM: a Versatile and Accurate Monocular SLAM System by R. Mur-Artal, J. M. M. Montiel, and J. D. Tardos
- Double Window Optimisation for Constant Time Visual SLAM by H. Strasdat, A. J. Davison J.M.M. Montielb, K. Konolige
- The Role of Wide Baseline Stereo in the Deep Learning World by Dmytro Mishkin
- To Learn or Not to Learn: Visual Localization from Essential Matrices by Qunjie Zhou, Torsten Sattler, Marc Pollefeys, Laura Leal-Taixe
- Awesome local-global descriptors repository
- Introduction to Feature Matching Using Neural Networks
Moreover, you may want to have a look at the OpenCV guide or tutorials.
- Pangolin
- g2opy
- ORBSLAM2
- SuperPointPretrainedNetwork
- Tfeat
- Image Matching Benchmark Baselines
- Hardnet
- GeoDesc
- SOSNet
- L2Net
- Log-polar descriptor
- D2-Net
- DELF
- Contextdesc
- LFNet
- R2D2
- BEBLID
- DISK
- Xfeat
- LightGlue
- Key.Net
- Twitchslam
- MonoVO
- Many thanks to Anathonic for adding the trajectory saving feature and for the comparison notebook: pySLAM vs ORB-SLAM3.
Many improvements and additional features are currently under development:
- loop closure
- relocalization
- map saving/loading
- modern DL matching algorithms
- object detection and semantic segmentation
- 3D dense reconstruction
- unified install procedure (single branch) for all OSs
- trajectory saving