ITS World Congress – Bordeaux 5-9 october 2015

Hi all!
In two weeks I am attending the ITS World congress in Bordeaux ( I’ll be presenting 3 papers in the scientific and technical sessions, but also be present at the Vicomtech’s booth F2.

We will be demonstrating our latest vision technologies, including the latest road model:

and the person drowsiness detection:

See you there!


This entry was posted in ADAS and tagged , . Bookmark the permalink.

4 Responses to ITS World Congress – Bordeaux 5-9 october 2015

  1. Selvin Jayakumar says:

    Dear Marco,
    Hope you are doing fine. I am currently working on a project to implement a lane and pedestrian detection system in an autonomous car. I have a query regarding the software we are intending to use. We are a bit skeptic about the use of OpenCV for this application since it is a bulky library and were wondering if OpenCV is a good option to consider when implementing a real-time system.
    Hope I could express the question in an understandable manner.

    Thank you.

    • Hi Selvin,
      OpenCV is probably the most appropriate tool to start doing things. You have plenty of algorithms in basically all sub-disciplines related to computer vision. And that makes OpenCV a need when you need something working, and in short time. Normally, for computer vision scientists, which aim at creating new algorithms, or comparing their methods against others, Matlab is even better, because of the greater amount of packages for statistics, image processing, etc.
      In my experience, we directly go for OpenCV-aided developments, where we use what we need from OpenCV and create our own C++ code for additional functions and the rest of application functionalities (e.g. persistence, visualization, communications, etc).
      For the case of lane and pedestrian detetion, I believe you can take advantage of OpenCV cv::Mat structures, try the detectors and classifiers (cv::SVM, also Adaboost, Cascade and Neural Networks), and jump into the OpenCV 3.0.0 T-API, which can be a great advantage if your HW support GPU acceleration.
      Of course, OpenCV is not the only one around. Recently, Dlib has grown a lot and contains a lot of interesting functions.
      At the end, you will need to find a functional system, using OpenCV or not, with clear input-output interfaces, and then optimize it into your target HW so that it runs real-time.
      In many cases, a thorough re-thinking of your entire pipeline will make you discover bottlenecks, redundant operations or information losses that can be prevented without really caring if implemented using OpenCV or other libs.

      Hope this helps!


      • Selvin Jayakumar says:

        Hey Marcos!

        Thank you for your reply! I will look into the things you said.
        I have a follow up question. If I had memory constrains and cannot afford to load the entire OpenCV library(which I assume is in GigaBytes) is there an alternate to this problem? Or is this not an issue at all? Our end product might be implemented on an embedded system which might not have enough memory to store the entire OpenCV library. This is where my major concern lies.

        Thank you.

      • Hi Selvin,

        Apologies for answering this late. I’ve been away of the blog for some time.
        Regarding your question, I believe that if you build OpenCV as dynamic libraries (.so in Linux) you won’t have problems for loading it in runtime. I don’t have the numbers, but I think it is in MegaBytes. We’ve used OpenCV in embedded platforms and the real problem are the models (the XML files) which are way larger.
        I also think that OpenCV 3 allows to compact everything into a single library, opencv_world, but I have not tried it.
        Good luck!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s