Hi! What we use is Viper (http://viper-toolkit.sourceforge.net/) because it's simple but complete enough for spatio-temporal event annotation in videos. LabelMe is another approach (http://labelme.csail.mit.edu/Release3.0/).

Hi,

Did you look at the Computer Vision web site at Carnegie Mellon University. There is a long list iof software. I recognise that most of it is rather old. There are functions infor this in OpenCV and I imagine that there wikll be Java apllets fior ImageJ. Might also be worth checking the libraries for Aforge.Net

We are developing model based software for tracking pedestrians. It works very well and we should finish a draft paper this week. The software is not packaged for release but we might b e able to make it available.

Regards,

David Pycock, University of Birmingham, School of EECE

For simple tasks, FIJI (http://fiji.sc/Fiji) is a very good tool. Permits some analysis and exporting results to an excel table. It is free and open-source. There are a lot of modules and an active community. It works on all platforms (Windows, OSX and Linux). A dissadvantage is that is is mainly biased to Biosciences, but even though I found it very useful. You can also find manuals and examples through the web.

Regarding more sophisticated tools, I recommend Matlab or any of its clones (Octave, Scilab). Libraries are really good and the users community is huge and really active.

I hope this helps you.

We have a tool available that makes it easy to mark tracked objects in videos using mouse hovering while progressing gradually through video frames. A bounding box follows the mouse cursor while the other hand progress the frames via keyboard. The tool allows intuitive, fast and precise tracking of moving objects. You can also change size of the bounding box via scroll wheel.

We also published a dataset of tracked human heads in consumer videos, you can use it to benchmark your tracking algorithms in various settings (indoor/outdoor, attentive/inattentive people) and recording sources (HD camcorder/mobile video). We used MPEG-7 as a data format.

Here's the publication that describes the tool and the dataset (published @ ACM Multimedia Systems 2013): https://www.researchgate.net/publication/257641093_Consumer_video_dataset_with_marked_head_trajectories

The Ground Truth Editor tool can be downloaded from the same place as the "Consumer"/User Generated Video dataset (at the bottom of the page):

http://www.mediateam.oulu.fi/downloads/MTUGV/

Hope you find this useful!

Conference Paper Consumer video dataset with marked head trajectories

Hi Martin,

I usually use the following combination:

- Viper tool (http://viper-toolkit.sourceforge.net/) in order to annotate the videos. It provides a file in a standard VIPER XML format, which can be used by the following framework;

- USF-DATE scoring tool (more details can be found in: "

Framework for Performance Evaluation of Face, Text, and Vehicle Detection and Tracking in Video: Data, Metrics, and Protocol" -Kasturi et al. ), which is able to compute standard indices such as MOTA, MOTP, ATA, SFDA and so on, as weel as false positive, id-switch etc.

Hello,

If your are looking for open source environment try opencv ...

I have never find common tools to perform evaluation of a method. In my case, I used a ad-hoc statistical approach. (If you are interested : A robust videogrametric method for the velocimetry of wind-induced motion in trees. Barbacci & al. Agricultural and Forest Meteorology. about to be available online)

Sincerly

Hi, VATIC looks very promising http://web.mit.edu/vondrick/vatic/

I've not tryied it for the moment so I cannot say more than that :)

Best regards,

ll.

Hi Martin,

We have developed a software (Sensarea) that can be used for annotating. Though I understand it is not sufficient for you, it is very straightforward and easy to use. You are welcome to give a look at a short demo at http://www.youtube.com/watch?feature=player_embedded&v=5kcBU0haACI

Dear Martin,

we have developed a tool (PerLa) (see http://groups.inf.ed.ac.uk/f4k/PAPERS/MTAP-Perla.pdf) which overcomes the difficulties of Viper (to me it is rather complex to use) and the ones developed at the MIT. At the moment the system is populated with our videos, but we can provide you the source code o create an account to upload your own videos. An interesting aspect of PerLa is how it supports the labeling of object trajectories by creating window chains, create chain with the desired number of consecutive frames.

If you are interested please contact me.

Regards,

Concetto

Dear Martin,

As a newbie using this kind of tools, let me recommend ELAN from Max Planck Institute for PsychoLingeuistics. Very handy and easy to download and install. used it for a EU project.

Best of luck

We use Noldus Observer (www.noldus.com, licensed) designed for experimenting in psychology. It works well and it also supports some semantics and consequently time sequence analysis.

But to be honest, a simple tool one implements for instance in Matlab (a couple of hours work) using Matlab guide does the job as well.

Look at the proceedings of the Workshops: http://vigta2012.dieei.unict.it/, http://vigta2013.dieei.unict.it/

You can check out this paper:

https://www.researchgate.net/publication/257227396_An_innovative_web-based_collaborative_platform_for_video_annotation

It's a very interesting idea, and maybe you can get in contact with authors.

Article An innovative web-based collaborative platform for video annotation

Hi Martin. I am the first author of "changedetection.net: A new change detection benchmark dataset". You may find my open source annotation tool interesting https://bitbucket.org/nilgoyyou/groundtruther. And you may want to use our online evaluation tool http://www.changedetection.net/. There's already a lot of methods listed there, it would be nice to have more!

Reading your requirements, I think it's exactly what you ask for. Maybe I'm wrong. At least you know that they exist now. Have a nice day.

The IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE journal publication, "Framework for Performance Evaluation of Face, Text, and Vehicle Detection and Tracking in Video: Data, Metrics, and Protocol", February 2009 (vol. 31 no. 2), Kasturi et.al, has an excellent open source evaluation tool for CLEAR and VACE metrics. It can be found at : http://doi.ieeecomputersociety.

org/10.1109/TPAMI.2008.57

I tend to use ViPER for annotating: http://viper-toolkit.sourceforge.net/products/gt/

Viper is a good tool even if it is an overkill for some applications.

Simple: OpenCV! :)

Check out the vatic tool from MIT.

http://web.mit.edu/vondrick/vatic/

Contact the lead author of the following paper by Surrey Space Centre UK - (https://www.researchgate.net/publication/239968761_Quasi-thematic_feature_detection_and_tracking_for_future_rover_long-distance_autonomous_navigation). He developed a similar tool recently and can help you customize it to your needs.

Conference Paper Quasi-thematic feature detection and tracking for future rov...

I have found good evaluation methods for tracking in:Black, J., Ellis, T., Rosin, P., "A Novel Method for Video Tracking Performance Evaluation," In Joint IEEE Int. Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance (VS-PETS}, pp. 125-132, 2003.

A good benchmark with evaluation data/code and server

http://www.cvlibs.net/datasets/kitti/

A popular annotation tool

https://code.google.com/p/labelme-video/

I tend to use the ViPER toolkit, though it it starting to show its age a bit. All the source code is available though, so it could be updated if needed.

http://viper-toolkit.sourceforge.net/

Dear Martin, my company has developed The Observer XT for this purpose (www.noldus.com/observer). It offers design of annotation schemes, data collection, visualization, reliability analysis (to test automated systems against ground truth), and more, but it's not open source. I can put you in touch with users in Denmark if you wish. Cheers, Lucas

Thank you all for your answers! I have looked at some of the suggestions and will go through them successively. Since video object tracking is a popular research field I was hoping that there would be something defacto standard or similar that I had missed. But judging from the diversity of the suggestions here it seems there is no standard way of doing this. Many people seems to do something by themselves ad hoc, I could as well but I think it would be better to use something more standard, robust, developed etc. To be continued.

ViPER, yes this tool is exactly what I'm after in functionality. We have tried it some years ago with some success, but the latest version is from 2005 and no signs of maintenance or updates since. It is dependent on Java 5 (if I remember correctly) which is an old version of Java. This is the drawback. Not impossible but cumbersome to need to have an old Java version installed on the particular machine, eventually switching between Java versions etc. Might also be some bugs with the GUI. Those of you here that use ViPER, how do you handle these things?

Maybe this answer is outdated, but a possible framework for you would be www.votchallenge.net/. It allows nice evaluation of tracking performance and comparison to other approaches.

Tom

Thank you for the advice on votchallenge.net and motchallenge.net, those are nice benchmarks but they don't have tools. After a lot of searching I did not find one tool that did both annotation and evaluation as ViPER does. But since ViPER is so old and cumbersome to maintain I have ceased to us it. There are many tools that do either annotation or evaluation so finally settled for using a combination instead of a one-stop-shop. For annotation I tried VATIC which is nice but has some annoying bugs that has not been fixed. For evaluation I tried pymot for generating statistics (mota/motp etc) which works well.

hi martin

i have a question: i have a DCB model in fracture mechanics(you can find it by search).i captured a film,16 seconds. now i want to calculate the speed of crack tip.can you explain for me,how can i do that?

thanks.

We have developed a video annotator that is a free, standalone, cross-platform application written in C++.  It supports bounding boxes, lines, and dots.  Annotations are saved in the widely supported JSON format.  FFMPEG is used to decode your videos, so it should work with most codecs.  Most importantly, we have gone to a lot of trouble to ensure frame accurate annotations, so even if you seek or step backwards the frame count associated with your annotations will remain accurate.  The only catch is that it was originally written with marine biology researchers in mind, so you will see categories referred to as "species".  Making the tool more general is on our to do list.  If anyone is interested I can work on a release, it is sort of a pain to build on Windows but less so on Mac or Linux. 

https://github.com/BGWoodward/FishAnnotator

Try CVAT (https://github.com/opencv/cvat).

Computer Vision Annotation Tool (CVAT) is a web-based tool which helps to annotate video and images for Computer Vision algorithms. It was inspired by Vatic free, online, interactive video annotation tool. CVAT has many powerful features: interpolation of bounding boxes between key frames, automatic annotation using TensorFlow OD API, shortcuts for most of critical actions, dashboard with a list of annotation tasks, LDAP and basic authorization, etc… It was created for and used by a professional data annotation team. UX and UI were optimized especially for computer vision tasks.

have you tried this one, http://darkpgmr.tistory.com/16 ?

a demo video at https://www.youtube.com/watch?v=vbydG78Al8s

hope this information useful.

Try using EVA annotation tool. https://github.com/Ericsson/eva ,this has an excellent tracking function. you mark the object in only 1 frame and rest/many of the frames are automatically annotated. This also has lock unlock feature to help annotate faster and more number of objects in each frame. This supports video upload or image data sets.

Best thing, completely Free/open source !