OpenTLD

Yolo and ccpmt
Yolo and Rolo tracking exceeds cppmt but would need an expensive Nvidia Jetson for realtime tracking. Cppmt in contrast will be able to run on Raspberry pi or DSP. Yolo draws a bounding box and outputs the coordinates to a text file. Use this text file to in turn automatically draw a bounding box for cppmt as opposed to manual drawing. Yolo's computational heavy resource usage means that for low power consumption deployment it must be ported to FpGa or DSP, instead of using an expensive Nvidia Jetson. This allows multiple Tracked vehicle turrets fitted with raspberry to get a coordinated image back from the base PC TitanX(over wifi MeshNetworking) or Jetson  processing 30 frames per second. https://www.youtube.com/watch?v=3iCifD8gZ0Q INTEL This short training introduces the high level concept of machine learning, focusing on Convolutional Neural Networks and explains the benefits of using an FPGA in these applications. FPGA can do 600 img/s at 31watts just on CNN, on Yolo it will be orders of magnitude faster. This allows for realtime behaviour, people, cars detection without having to stream large amounts of bandwidth to a central base. We would be able to secure an entire city with just cellular connectivity.

cppmt
http://www.gnebehay.com/cmt (OpenTLD) Clustering of Static-Adaptive Correspondences for Deformable Object Tracking (CMT) is an award-winning object tracking algorithm, initially published under the name Consensus-based Tracking and Matching of Keypoints for Object Tracking at the Winter Conference on Applications of Computer Vision 2014, where it received the Best Paper Award. A more detailed paper was published at the Conference on Computer Vision and Pattern Recognition 2015. CMT is able to track a wide variety of object classes in a multitude of scenes without the need of adapting the algorithm to the concrete scenario in any way. Experiments have shown that CMT is able to achieve excellent results on a dataset that is as large as 77 sequences. A C++ implementation (CppMT) is freely available under the BSD license, meaning that you can basically do with the code whatever you want. Additionally, the original Python research code is still available for reference.
 * https://github.com/gnebehay/CppMT Newer algorithm preferred over Kalal's previous OpenTLD.

CMT
https://github.com/gnebehay/CMT used by Github

ROS platform

 * http://diydrones.com/profiles/blogs/computer-vision-for-drones-using-flytapis-part-2-object-tracking opentld on ROS robotics.
 * http://docs.flytbase.com/
 * http://flytbase.com/

Object tracking with opentld
Opentld can do face recognition in videos.


 * http://www.youtube.com/watch?feature=player_embedded&v=1GhNXHCQGsM
 * http://www.youtube.com/watch?v=W2qR60hrD2w&feature=player_embedded
 * http://www.youtube.com/watch?feature=player_embedded&v=tKXX3A2WIjs
 * http://www.youtube.com/watch?v=qvcyK4ZMKbM&feature=relmfu Human face tracking
 * http://www.youtube.com/watch?v=4EQohE-Wyl0&feature=relmfu Motorbike tracking
 * http://www.youtube.com/watch?v=rwGhWJ3wg6Q&feature=related Car tracking, ignore pedestrians PREDATOR opentld
 * http://www.youtube.com/watch?v=7YacY833dds&feature=related Tld tracking demonstration
 * http://www.youtube.com/watch?v=e5oOIS1WJEc&feature=related opentld handball tracking test
 * http://www.youtube.com/watch?v=9eZXlUPuCGY&feature=related Mobile opentld
 * http://www.youtube.com/watch?v=z9G8DT-GkyQ&feature=related Use opentld to identify strange faces, thus if you are surrounded by criminals the PepperBall gun turret will avoid shooting you and only the intruders. OpenTLD can lock on and track an intruder allowing a mobile robot to shoot with air gun. See http://visionlabs.cl/blog/?p=5 or

Liuliu fork

 * http://libccv.org/post/ccv-now-has-a-state-of-art-tracking-algorithm/
 * https://github.com/liuliu/ccv

Links
http://groups.google.com/group/opentld Get membership, main discussion page

https://github.com/gnebehay/OpenTLD

https://github.com/arthurv/OpenTLD

https://github.com/alantrrs/OpenTLD

http://www.tldvision.com

http://info.ee.surrey.ac.uk/Personal/Z.Kalal/tld.html

https://github.com/anikita/openTLD

frk
https://groups.google.com/group/opentld/browse_frm/thread/1631efba8ef2fd62

Gnebehay fork
http://jayrambhia.wordpress.com/2012/05/25/opentld-georg-nebehays-version-of-opentld-in-opencv-2-3-1-and-towel-day/

https://github.com/jmfs/OpenTLD/tree/gnebehay_videoinput forked from arthurv

I have a similar issue. I have it working such that it runs but no > camera feed is showing up in the pop-up window. I'm using Georg > Nebehay's version. > Does anyone know how to fix this? Thanks!

Well, luckily for you, I've been working on just that! OpenCV's camera input doesn't work with a lot of cameras, so I've modified Georg's version to get the video from videoInput instead. You can find the code here: https://github.com/jmfs/OpenTLD/tree/gnebehay_videoinput. Instructions are the same, except that you may have to edit the opentld project's properties, since I haven't figured out how to make that work via CMake. In the Linker/input tab, put "atlthunk.lib;libcmt.lib" into "Ignore specific default libraries". -- João Silva

> Just one more quick question. Does this still work if I wanted to use > a video from an external source (streaming via wifi) or from a video > saved on my computer? If so, how should I edit the code to accommodate > that modification?

For the locally saved video, the support is there, but OpenCV may not support the video codecs. Check the readme for the appropriate command line options and try it out. As for the streaming, you'd have to modify the code to do it. The code's pretty modular, so pretty much everything you need to modify is in imAcq.cpp and imAcq.h. -- João Silva

Forks
My github repo (forked from Alan's) can be found here: >>> https://github.com/arthurv/OpenTLD >>> Alan's repo can be found in:https://github.com/alantrrs/OpenTLD. We >>> merge changes in these branches periodically.

post1
https://groups.google.com/group/opentld/browse_frm/thread/1631efba8ef2fd62

https://groups.google.com/group/opentld/browse_frm/thread/1631efba8ef2fd62

I got the purely C++ working (without MATLAB) from the https://github.com/gnebehay/OpenTLD Just grab all the code, then get it works with Visual Studio 2010: Below is the steps : 1)Download the code from link : https://github.com/gnebehay/OpenTLD/zipball/master 2)Then Download the Open CV 2.2 and CMake 2.8 3)When installing the Open CV2.2 make sure with "Add OpenCV to the system PATH for all users" in Install Options and also for the CMake in the Install Options. 4)After you have installed the CMake, Click the CMake and then setting the path: >Where is the source code : (Choose the folder that you have download from the https://github.com/gnebehay/OpenTLD/zipball/master) > Where to build the binaries : (Make the build folder inside the folder) > Configure to Visual Studio 2010, then Generate 5)After you'v done the CMake, select the folder then choose OPENTLD.sln (solution file) which makes you open the Visual Studio 2010 automatically. 6)Click 'View' then choose 'Property Manager' 7) Then click tld > Release|Win32 > Microsoft.Cpp.Win32.user. 8) Then You do right click on it, choose Properties 9) In Common Properties, choose VC++ Directories   Include directories : C:\Program Files\OpenCV2.2\include; C: \Program Files\OpenCV2.2\include\opencv; C:\Program Files \OpenCV2.2\include\open2    Library Directories :C:\Program Files\OpenCV2.2\lib 10) Then Ok 11) Then click 'View' to choose the 'Solution Explorer', then build the solution 12) Don't click F5, scroll the option, make sure you choose 'Release' 13) Then F5 14) Will not run the program but succeed, no failed 15) Close the Visual Studio 2010 16) Then open the folder build > src > Release > tld.exe Hope you can follow this steps carefully Thank you Library Directories : On Apr 13, 8:03 pm, sameer  wrote:

Other tracking solutions
http://www.youtube.com/watch?v=LNaL58GRou4&feature=related OpenCV

Developers
https://groups.google.com/group/opentld/browse_frm/thread/5286ba1a77fed2ed

Redwood Center for Theoretical Neuroscience University of California at Berkeley 567 Evans Hall, MC# 3198 Berkeley, CA 94720-3198, United States School of Computer Science and Technology Harbin Institute of Technology 92 Xidazhi Street, Harbin 150001, China Phone: 510-502-8557 / +86-451-86416485 http://redwood.berkeley.edu/spzhang/

Other notes
OpenTLD android Federico Pernici