Welcome to Py2vision’s documentation!

Here you can see every detail about py2vision package which is a tool to do object positioning with stereo vision and object detection, this package can be used in Robotics and Computer Vision.

Installation

Open a console and write this: pip install py2vision

Important

If you already have the opencv package installed, you have to uninstall it with pip uninstall opencv-python and install opencv-contrib-python version 4.6.0.66.

What could you archieve with py2vision?

  1. It will allow you to obtain disparity maps using SGBM algorithm.
  2. Get distances in the 3D space.
  3. Measure objects dimensions and more.
  4. Train your own object detection model and infer with it.
  5. Combine detection and stereo vision to get homogeneous coordinates of objects im an scene.
https://github.com/corvus96/PyTwoVision/blob/master/source/doc_static/disparity_map_example.jpg?raw=true

Example of a disparity map

https://github.com/corvus96/PyTwoVision/blob/master/source/doc_static/position_example.jpg?raw=true

Example of positioning of objects

Modules

Dependencies

  • numpy == 1.21.5
  • tensorflow == 2.8.0
  • opencv-contrib-python==4.6.0.66
  • wget == 3.2
  • pandas
  • pyyaml
  • h5py

References

[1]
  1. Wheatstone, «Contributions to the Physiology of Vision.», Proceedings

of the Royal Society of London, vol. 4, 0 1837.

[2]
    1. Martín, Técnicas de visión estereoscópica para determinar la estruc-

tura tridimensional de la escena Proyecto, Madrid, 2010.

[3]
  1. Dembys, Y. Gao, A. Shafiekhani y G. Desouza, Object Detection and Pose

Estimation Using CNN in Embedded Hardware for Assistive Technology, oct. de 2019.

[4]
  1. Königshof, N. O. Salscheider y C. Stiller, «Realtime 3D Object Detection

for Automated Driving Using Stereo Vision and Semantic Information», en 2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019, págs. 1405-1410. doi: 10.1109/ITSC.2019.8917330.

[5]
    1. Barnard y M. A. Fischler, «Computational Stereo», ACM Computing

Surveys (CSUR), vol. 14, 4 1982, issn: 15577341. doi: 10.1145/356893. 356896.

[6]
  1. Torres, J. Pomares, P. Gil, S. T. Puente y R. Aracil, Robots y Sistemas

Sensoriales. Madrid: Pearson Education, 2002, págs. 69-94.

[7]
  1. Zhang, «A flexible new technique for camera calibration», IEEE Transac-

tions on Pattern Analysis and Machine Intelligence, vol. 22, 11 2000, issn: 01628828. doi: 10.1109/34.888718.

[8]
    1. Hartley, «Theory and practice of projective rectification», International

Journal of Computer Vision, vol. 35, 2 1999, issn: 09205691. doi: 10.1023/ A:1008115206617.

[9]
  1. Scharstein y R. Szeliski, «A taxonomy and evaluation of dense two-frame

stereo correspondence algorithms», International Journal of Computer Vi- sion, vol. 47, 1-3 2002, issn: 09205691. doi: 10.1023/A:1014573219977.

[10]
  1. Szeliski, Computer Vision : Algorithms and Applications 2nd edition.
[11]
  1. Dai y W. Huang, A-TVSNet: Aggregated Two-View Stereo Network for

Multi-View Stereo Depth Estimation, 2020. arXiv: 2003.00711 [cs.CV].

[12]CS231n Convolutional Neural Networks for Visual Recognition, Accesado el

12, de febrero de 2021. dirección: https://cs231n.github.io/convolutional- networks/.

[13]
  1. Kar, Mastering Computer Vision with TensorFlow 2.x. 2020.
[14]
  1. He, X. Zhang, S. Ren y J. Sun, Identity Mappings in Deep Residual

Networks, 2016. doi: 10 . 48550 / ARXIV . 1603 . 05027. dirección: https : //arxiv.org/abs/1603.05027.

[15]
  1. Atienza, Advanced Deep Learning with Keras: Apply deep learning tech-

niques, autoencoders, GANs, variational autoencoders, deep reinforcement learning, policy gradients, and more. 2018.

[16]
  1. Long, E. Shelhamer y T. Darrell, Fully Convolutional Networks for Se-

mantic Segmentation, 2014. doi: 10.48550/ARXIV.1411.4038. dirección: https://arxiv.org/abs/1411.4038.

[17]——, StereoPi V2 quick start guide, 2014. dirección: https://wiki.stereopi.

com/index.php?title=StereoPi_v2_Quick_Start_Guide.

[18]
    1. Adrian Kaehler, Learning OpenCV 3: computer vision in C++ with

the OpenCV library. California, USA: O’Reilly Media, Inc., 2016.

[19]T.-Y. Lin, P. Dollár, R. Girshick, K. He, B. Hariharan y S. Belongie, Feature

Pyramid Networks for Object Detection, 2016. doi: 10.48550/ARXIV.1612. 03144. dirección: https://arxiv.org/abs/1612.03144.

[20]
  1. Redmon y A. Farhadi, «YOLOv3: An Incremental Improvement», arXiv,
[21]
  1. Everingham, L. Van Gool, C. K. I. Williams, J. Winn y A. Zisserman,

The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results, http://www.pascal-network.org/challenges/VOC/voc2012/workshop/index.html.