NVIDIA CEO Jensen Huang is famous for his keynote addresses at the company’s GPU Technology Conference.
Jensen delivered this year’s keynote yesterday, and the focus was deep learning and artificial intelligence. NVIDIA GPUs are critical for training deep neural networks, so NVIDIA is fast becoming as much an AI company as a gaming company.
Engadget created a super-fast mashup of Jensen’s keynote, if you only have 13 minutes to catch the highlights. In particular, check out the autonomous vehicle announcements around NVIDIA Drive and Guardian Angel.
Ricardo’s lane-finding pipeline works amazing well on challenge video for the Advanced Lane Finding Project. He has an incredibly thorough rundown of his pipeline: calibration, undistortion, color transforms, perspective transform, lane detection, curvature, and unwarping:
“Once you know where the lines are in one frame of video, you can do a highly targeted search for them in the next frame. This is equivalent to using a customized region of interest for each frame of video, which helps to track the lanes through sharp curves and tricky conditions.”
I just met Yazeed yesterday at NVIDIA’s GPU Technology Conference, and then I found his blog post today. He loves GPUs and deep learning so much he flew all the way from Saudi Arabia for the conference!
“For me it was a gambling to buy 1200$ Titan Xp which is just relased 17 houres ago — when I bought it — with a promise from NVIDIA to support macOS and I don’t have the Thunderbolt3 port which is the supported port for eGPU. So, I said like Richard Branson said “Screw It, Let’s Do It” and it works like a charm. without further ado, let’s dive in.”
A while back Mez published his results on behavioral coning with SqueezeNet, but he’s back with a super-enthusiastic blog post on his network. Only 52 paramters and 6 second epochs on a CPU!
“One good rule of thumb I developed from this project is to try and reduce the number of variables you are tuning to gain better results faster.”
Arsen has a great writeup on using a neural network to calculate both steering and throttle values for the Behavioral Cloning Project. Also, he uses early stopping to prevent overfitting the data.
“The validation set helped determine if the model was over or under fitting. I used EarlyStopping (utils.py line 299) to stop training when validation mse has stopped improving.”
JC’s writeup of his Behavioral Cloning Project covers a really important topic — how to debug, or at least visualize, what’s going on inside a neural network.
“During the process of training, I felt very uneasy as it is almost like a blackbox. Whenever the model failed to proceed at a certain spot, it is very hard to tell what went wrong. Although my model passed both tracks, the process of try and error and meddle around with different combination of configurations is quite frustrating.”
Mobileye banks on mapping revenue. There have been very few companies able to supply the high-definition maps that autonomous vehicles need. Mobileye plans to be one of them.
K-City will dwarf M-City. Korea is building an urban testing environment for self-driving cars. It will be almost three times bigger than the environment that Ford and the University of Michigan built.
Last week I traveled with colleagues to Sonoma Raceway for Safe Driver Training, a mandatory class for autonomous vehicle operators.
The class itself is not oriented around autonomous vehicles, but rather how to anticipate and evade dangerous situations on the road. The logic of requiring this class of autonomous vehicle operators, I suppose, is that if you have to take over the vehicle in an emergency, hopefully you are able to anticipate and evade a collision.
The biggest lessons are to look as far ahead as possible. Sit lower in the vehicle and raise your eyes toward the horizon. Then, when performing an evasive maneuver, lock your eyes on where you want the vehicle to go. Or, as the instructors say, “Keep your eyes on safety.”
The class was a lot of fun. Several of the exercises involved negotiating tight turns at high speed, just like if an obstacle popped out at the last minute. Other exercises required us to spin out the vehicle in a tight turn, then regain control and proceed through a gate.
Here’s a practice run for a tight turn exercises — the procedure gets tougher when they don’t tell you which way to turn until the last second:
Since the program is held at Sonoma Raceway, there all sorts of cool racecars around.
Nearly all of my savings are in various index funds, but I do own stock in one, single individual company: Berkshire Hathaway.
It’s mostly for sentimental reasons. I went to Omaha a couple of times during business school: once for the Berkshire annual conference (“Woodstock for Capitalists”) and once to meet the Oracle himself, as part of a school trip.
I’ve known for a while that autonomous vehicles would hurt insurance, which is one big part of Berkshire’s business. The logic is that insurance companies only exist because drivers need to insure themselves against the costs of accidents. If accidents diminish, the need for insurance diminishes.
But a question at this year’s annual meeting pointed out that another big part of Berkshire’s business is highly vulnerable to autonomous vehicles: railroads.
Berkshire purchased the Burlington Northern Santa Fe (BNSF) railroad for $26.5 million in 2010 and it’s been a good investment.
That investment will come under intense pressure from self-driving trucks, however. Once trucks can operate nearly constantly, without the cost or physical limitations of a driver, the cost advantage of transportation by rail will diminish, or maybe even disappear completely.
If you’ll be there, please stop by to say hello. There will be a car display, plus instructors and students talking about the Self-Driving Car Nanodegree Program.
There are still tickets left to the conference if you’d like to register! If you’re a Udacity student, email me (david.silver@udacity.com) for the student discount code.
Here are stories from Udacity students about what they wish they knew in the past, what they’re doing in the present, and what they hope to do in the future!
A self-managed team of Udacity students from around the world competed at the Self-Racing Cars event in California last month. They put in a ton of work and here’s what they learned:
“On February 15, Udacity selected the group of 18 talented engineers (out of hundreds of applicants) to form the Self-Racing Cars team. Our team was composed of individuals with largely varying backgrounds from all over the world, with the commonalities that we were all enrolled in the Udacity Self-Driving Car Nanodegree program, and extremely passionate about autonomous vehicles. The team was given six weeks to develop the software to drive an autonomous vehicle around the track at Thunderhill Raceway for the Self-Racing Cars event.”
Daniel, bless his heart, put together a terrific list of tips and tricks for Term 1 of the Nanodegree Program. He would know, after having mentored over 40 students!
“If I could send myself a note back in time to 6 months ago, I would probably find something more valuable than sending myself mentorship tips for Term 1 of the Udacity Self-Driving Car Nanodegree. That being said, I would have wanted to know these points soon after being accepted into the selective inaugural cohort in October 2016. I have mentored over 40 students after having some success in the SDC Nanodegree myself, and this post will reveal the pointers that have been most relevant to my mentees.”
Check out how Eirik built these super-cool lane line GIFs!
“This project is the first among several projects in the Self Driving Car Engineer program at Udacity. Here we learn cutting edge technology equipping us with the tools for a career in the field of Self Driving Cars. Udacity calls it a “Nanodegree”, but it lasts over 9 months and with all the hours I am putting into this, it really becomes a full education for me. So that brings me to this project, which was so much fun I just had to stop myself, I could go on forever, there is always something to do better, and so much good Inspiration!”
Riccardo provides a terrific and thorough walkthrough of his vehicle detection project. I especially like the experiments he ran with color spaces and histograms.
“First, we identify and extract the features from the image, and then use it to train a classifier. Next, we execute a window search on the image, on each frame from the video stream, to reliably identify and classify the vehicles. Finally, we must deal with false positives and estimate a bounding box for vehicles detected.”
I love Michael’s story of leaving his Big Four accounting job to become a self-driving car engineer. He completed Term 1!
“Luckily in Silicon Valley many people are more focused on what you can do than simply how many years you’ve been doing something. For Udacity’s part, they’ve provided me with a mentor and lots of career content, as well as access to events with some great hiring partners, that also give me great hope that I’ll be able to make the jump to working directly on self-driving cars.”
Nanalyze has a brief roundup of six vehicle-to-vehicle (V2V) startups to watch. What’s striking to me is how many of them have been around for quite a while — almost a decade in some cases:
Vehicle-to-vehicle communication is really exciting — imagine the hypothetical world with no traffic lights, because cars communicate with each other and weave through intersections.
This hypothetical future, though, might be a long ways out. V2V suffers right now from being at the losing end of a network effect — because almost nobody has V2V technology in their cars, it’s not particularly valuable to have V2V technology in your own car.
On a more immediately feasible note, he also says, “November or December of this year, we [Tesla] should be able to go from a parking lot in California to a parking lot in New York, no controls touched at any point during the entire journey,”
That’s pretty awesome, although the catch with these things is how generalizable the solution is.
If the demonstration works only from one very specific parking lot in New York, to another very specific parking lot in California, and only over one precise cross-country route, that’s impressive but not groundbreaking. Delphi actually did something like that a few years ago.
If, on the other hand, Tesla builds a system that can drive over a wide variety of routes, that will be a huge step toward Level 5.
The cars are being tested in the Phoenix suburbs, and Waymo published a cute video with one of the families that has been testing the car.
To watch the video, it appears that the vehicle does not have a safety driver, although perhaps the family members are trained to operate self-driving cars in an emergency.
Now that Waymo is bringing the program out of stealth, it’s recruiting more families to try out the service, so if you live in the Phoenix area, you should apply!
This is a particularly interesting announcement, because speculation has been rampant about what Waymo’s next move is with self-driving cars.
Google (Waymo’s parent company) has had self-driving cars spinning around Mountain View for years, with paid test drivers. The caution about putting real passengers in these cars caused a lot of people to question whether Google was going to give up a big lead to more aggressive companies like Tesla and Uber.
It’s good to see Google getting out there and ramping up it’s customer base. Here’s hoping the vehicles come to northern California soon.
David Silver 