Investigating the mysterious voice as a test case for a new (computational) idea of ​​voice design

In the video above, Lambert shows how the CV tool identifies the ball’s spin axis with a neon green dot and the seam position in other neon hues.

“I think that’s probably the most effective use of CV right now is to … get some metrics that I can’t get from Hawk-Eye,” Lambert said. “I’m sure you can imagine if my high school brother in Cincinnati threw a bull and he didn’t have a Trackman available, if we could get some pictures, get some measurements of what’s going on, we’d better get that process right from there.”

A computer vision system learns by analyzing thousands upon thousands of labeled images – sometimes even millions like Tesla’s early self-driving efforts – using convolutional neural networks to identify patterns and understand spatial sequences. This is a serious read.

Boddy and others at Driveline did tons of labeling, heavy lifting, to train the system – labeling seams, spin axis, and types of thousands of recorded contributions. The program is still learning, it is still getting better.

Lambert made nearly 50 throws earlier this month, learning the ball’s flight and the impact of each adjustment guided by feedback from Driveline’s real-world AI effort.

He couldn’t exactly replicate the pitch in one bullpen, but he was able to mimic some of its characteristics after just one hold and outputting a recommendation from our computer vision model and making some adjustments.

“What I was able to do again is that I was able to run to the side of the arm that he was going to throw,” said Lambert. “I couldn’t kill the spin efficiency to get the gyro action. I found it easier to create a change profile with the attachment (release) than to create a true gyro version of his slider. That was a lot of iterative process. It actually didn’t take that long to produce pitches with high velocity, on the arm side.”

Think what real pro and college pitchers and coaches can do with the tool?

That’s one application of a computer vision model: to help coaches and players understand how to start pitching.

Driveline director Connor White explains another major benefit of advanced design aided by deep learning.

“The speed of analysis is one of the most exciting things,” White said. “We want to keep those pens as a game. So, if you have to stop after each pitch and look at a bunch of metrics and contact the video, and the next thing you know it’s been a minute between pitches or more it’s kind of flowing breaks… Computer vision allows you to look at the observed versus supported (movement), to get closer to what’s happening on the ball.

“The speed at which these (developments) can be implemented is very exciting.”

Reducing the feedback loop, understanding what the voice is doing, is really exciting.

Our computer vision model is not a finished product, but it is already having results in our gyms.

Driveline pitching coach Grayson Liebhardt says it has already helped him as a coach.

“It’s a really useful tool,” Liebhardt said. “It’s still early in development but it helps us close the gap, and understand the shape of the seam without having access to the data that the professional organizations have… It gives us more context as to why the pitch may move in a certain way, or, how to improve the shape of the seam for certain movement profiles.

“Pitch physics is not completely solvable. There are many, subtle things that aren’t Magnus, like seam wake, and other variables that we may not even be aware of, that affect the movement of the ball,” Liebhardt said.

For example, Liebhardt notes that we know how seam wake affects ball flight but we can’t determine how much it affects movement in conjunction with other variables, something he notes that “we may not have figured out yet.”

We don’t know everything. And what’s more fun to see with a computer will lead to more understanding.

“These tools are very useful for using the information we already have,” he said of CV, “and gathering more information so we can learn more about pitch physics.”

One of the exciting things about real-world AI success is that they keep learning, they keep getting better.

“The cool part for me is being able to have an easy way to check seam orientation and spin axis,” Liebhardt said. “That’s just something, historically, you would (read) an Edgertronic camera and try to find it and guess where the spin axis is.”

Now, Liebhardt has a tool that cuts through most of the guesswork.

He shared this clip of another mysterious Imai-like voice, this one from Driveline host Tony Oreb.