Tip:
Highlight text to annotate it
X
Aaaarthuuur here!
Journalism intern, 27 years old, getting ready to meet some PhD students…
…people who are going to change my life!
(Bright Spots and Blurry Drops)
Alright, my friend, you’re about to come face to face with everything you’re not.
So, show yourself some respect.
Turn on your natural charisma…
…and stay calm!!
Ok, good view on the room.
You’re not early, not late.
Alright, let’s be daring – get those colors out!
Serenity…
Put him at ease, and you, you stay straight.
Complete transparency!
He’s here! All hands on deck!
Where’s my hot cocoa with whipped cream?!
I’m Raoul de Charette. I did 3, almost 4, years of research for my PhD,
in the Robotics Lab of MINES ParisTech.
I worked on vision algorithms, or what we usually call computer vision.
The goal was to develop new methods of vision, of image analysis, to be used in intelligent cars.
My studies had nothing to do with engineering.
I don’t have a scientific or engineering background, at all.
But the common theme is working with images.
In graphic arts school, I approached it from an information diffusion point of view.
Then, an artistic one, at the University of Paris 8.
And then from a technical point of view for my PhD.
I’m lucky that I get to travel.
These last few years I’ve been to Singapore, Malaysia, Cambodia, China…
All thanks to my PhD studies.
Another subject I worked on during my PhD was the detection of blurry drops.
Using a camera inside the vehicle, behind the windshield,
we try to detect drops on the windshield, while the camera is focused in the distance.
This could have different applications:
for improving other vision algorithms…
and automatically activated windshield wipers, for example.
Here, look. This drop looks blurry, because the camera is focused on infinity.
The red in this image represents zones detected as blurry. Probably zones containing blurry drops.
To take this analysis further,
we can also check the distribution of spatial frequencies.
For that, the literature tells us that a specific pattern should appear for blurry, out-of-focus zones.
Here, we see the result of a technique called cepstrum.
We’ve isolated the blocks that correspond to the pattern we’re looking for.
The two techniques we developed during my PhD let us form hypotheses about the presence of blurry drops in the image.
We can then prove or disprove these hypotheses with physical criteria, for example.
Another research subject I worked on is detection of traffic lights in an image –
again, in an image, from a camera inside the car.
We try to detect the bright spots in the image, the sources of light. For example, traffic lights.
Once we’ve detected all the light sources, including traffic lights,
we analyze the neighboring region
and see if it corresponds to a model we’ve identified previously.
We apply a template matching technique to this model.
“Bottom-up template matching” is the term.
And that lets us determine if these regions are traffic lights, or not.
Doing a PhD was something that had a big impact on me. It’s a great professional and personal experience.
Working in a lab with researchers who have really different ideas,
who work on really different subjects, is fascinating.
- In the future, I hope to continue in research. - Oh, yes?
Yes, in research, developing new projects. Probably in academic research.
Arthur: Well, listen, thanks a lot, Raoul.
Thank you.
Arthur: Take care.
You never know…See you somewhere down the road. Bye.
What a great guy.
I didn’t dare ask any questions. I let him talk.
It went so fast. I shouldn’t have been so shy.
He started opening up to me just when we had to go.
Raoul’s a really busy guy.
He’s exploring the world and taking part in its technical evolution.
His path is unusual, and yet so brilliant.
The desire to discover things… This passion for moving forward seems to have guided him.
Am I passionate…about moving forward?
Written and directed by Paul Garcia In the role of Arthur - Dimitri Pougnet The researcher - Raoul de Charette, as himself