APIDIS tracks players and the ball to select the best camera to use for each moment of the game.

Researchers at the Universite Catholique de Louvain (UCL) in Belgium have successfully developed software that can track basketball players and the ball as they race around the court. The project, Autonomous Production of Images Based on Distributed and Intelligent Sensing (APIDIS), can select between the video feed of several different cameras to find the most interesting footage at a given time. The software can even be told to show preference for a particular player, the star of the team perhaps, and choose the shots which include that player in the frame. It’s also been tested on football (soccer) matches with similar success. APIDIS may be used to reduce the need for human TV staff, making it cheaper to produce video coverage of games with a small audience (i.e. High School, small colleges). We have a video from New Scientist below, as well as some raw data footage, to demonstrate APIDIS in action. This technology could have major sports networks letting computers call the shots, and it heralds some important changes in journalism, surveillance, warfare, and the way we record our lives.

It seems like automation is really focused on journalism. We’ve already seen software that can write a decent baseball story. Statsheet, a sports statistics website, is planning on developing it’s own writer this summer. There’s a robot in Japan that will investigate, interview, and photograph until it finds a story worth publishing on the internet. Now, with APIDIS, computers are taking over the roll of directors and editors, finding the best shots and blending them together to form the visual narrative. How long before automation has come to synthesize every step of the reporting process? In the future, machine-run journalism may allow for a far-broader and cheaper style of reporting, allowing the creation of micro-news feeds tailored to audiences of a few hundred or less. Maybe even an audience of one. Our automated journalists could write stories, and create news footage personalized for each of us. Humans can’t compete with that kind of specificity. It looks like the robots’ first step towards global domination is control of the press.

Footage from APIDIS showing how a single player can be tracked across the court.

More videos of APIDIS footage of basketball and football matches can be found here, and the raw datasets can be found here.

APIDIS also has implications in surveillance and other forms of video monitoring. Software like Vitamin D Video (already on sale) allows for a user to have hours of surveillance footage reduced to a few key highlights for inspection. Now, we have the means to take that automation/evaluation process one step further. A human could never watch the live feeds from a 100 CCTV cameras at once and try to monitor them all for criminal activity. If APIDIS was adapted to that end, it likely could. Together, Vitamin D and APIDIS could take flawed and partially ineffective camera security systems and turn them into effective tools for reporting crime as it happens. As such real time surveillance becomes effective, it may also become preventative.

We also have to remember that military actions are becoming increasingly mechanized. Surveillance drones in the US Armed Forces provide soldiers with a nearly endless stream of video footage. As software continues to advance, it will reduce that footage to highlights that will inform and direct commanders in the field. If programs like Vitamin D and APIDIS become sophisticated enough they could increase the already incredible advantage that unmanned vehicles give to modern armies.

If those seem like heavy consequences for software meant to improve sports coverage, I should mention that UCL has always had higher goals for APIDIS. Sports journalism is one application, so is harnessing the video from mobile devices. APIDIS is part of the ICDSC Challenge, which seeks innovative uses of mobile camera feeds. Could we see APIDIS as a tool to take the thousands of camera phone videos at a concert to provide one movie of the event? The same could be done for protests, disasters, or your child’s birthday party. Sometime in the future it may be possible to run an application on your smart phone letting your video feed be combined with thousands or millions of others as a sort of video social networking. Journalism, surveillance, family movies all using the same kind of technology. Once computers behind the cameras are made smarter, almost anything is possible.

[screen capture credit: APIDIS]
[video credits: New Scientist, APIDIS]
[source: UCL, APIDIS, Paper Published for WIAMIS]

Computers can now write a baseball news article. America's pastime just went 21st century.

Sports writers are sharpening their pencils and maliciously eying their laptops as researchers in the Intelligent Informations Lab at Northwestern University announce that they’ve created a program that can write a convincing baseball news story without human help. Called Stats Monkey, the new computer software analyzes the box scores, and play by plays to automatically generate the news article. It highlights key players and clutch plays and will even write an appropriate headline and find a matching photo for a few player! Check out a sample story written by Stats Monkey after the break.

Not satisfied with worrying only baseball sports writers, developers say Stats Monkey could be adapted to any situation with a heavy helping of statistical data. Translation: it could work for every sport humans like to read about. Moreover, Stats Monkey could be adapted to write business stories, or conference updates, or other forms of professional journalism that rely heavily on numbers and analytics. Writing, it seems, is no longer immune from automation.

Stats Monkey basically combines two sections of code: the first uses baseball analysis algorithms to determine what actions in the game were important. The second part of code selects from several archetypal narrative arcs (the come from behind win, the landslide victory, etc) and adapts one to fit the stats of the game. It can even write from the point of view of either team. That’s more than can be said of some sports journalists out there.

Without further ado, check out the Stats Monkey story for the October 11th playoff game between the Angels and Red Sox:

BOSTON — Things looked bleak for the Angels when they trailed by two runs in the ninth inning, but Los Angeles recovered thanks to a key single from Vladimir Guerrero to pull out a 7-6 victory over the Boston Red Sox at Fenway Park on Sunday.
Guerrero drove in two Angels runners. He went 2-4 at the plate.
“When it comes down to honoring Nick Adenhart, and what happened in April in Anaheim, yes, it probably was the biggest hit (of my career),” Guerrero said. “Because I’m dedicating that to a former teammate, a guy that passed away.”
Guerrero has been good at the plate all season, especially in day games. During day games Guerrero has a .794 OPS. He has hit five home runs and driven in 13 runners in 26 games in day games.
After Chone Figgins walked, Bobby Abreu doubled and Torii Hunter was intentionally walked, the Angels were leading by one when Guerrero came to the plate against Jonathan Papelbon with two outs and the bases loaded in the ninth inning. He singled scoring Abreu from second and Figgins from third, which gave Angels the lead for good.
The Angels clinched the AL Division Series 3-0.
Angels starter Scott Kazmir struggled, allowing five runs in six innings, but the bullpen allowed only one runs and the offense banged out 11 hits to pick up the slack and secure the victory for the Angels.
J.D. Drew drove in two Red Sox runners. He went 1-4 at the plate.
Drew homered in the fourth inning scoring Mike Lowell.
“That felt like a big swing at the time,” said Drew. “I stayed inside the ball and put a good swing on it. I was definitely going to be ready to battle again tomorrow, but it didn’t work out.”
Drew has been excellent at the plate all season, especially in day games. During day games Drew has a .914 OPS. He has hit five home runs and driven in 17 runners in 36 games in day games.
Papelbon blew the game for Boston with a blown save. Papelbon allowed three runs on four hits in one inning.
Reliever Darren Oliver got the win for Los Angeles. He allowed no runs over one-third of an inning. The Los Angeles lefty struck out none, walked none and surrendered no hits.
Los Angeles closer Brian Fuentes got the final three outs to record the save.
Juan Rivera and Kendry Morales helped lead the Angels. They combined for three hits, three RBIs and one run scored.
Four relief pitchers finished off the game for Los Angeles. Jason Bulger faced four batters in relief out of the bullpen, while Kevin Jepsen managed to record two outs to aid the victory.

The news that Stats Monkey is even attempting to replace them should have writers everywhere gathering up powerful magnets and heading to Northwestern’s hard drives to extract some retribution. As it is, many journalists have taken a humorous spin on the situation (some quite defensive). That could be laughter in the face of death. It seems almost inevitable that at least some forms of writing will be automated, and sooner than expected. As such, I’d like to take the time to make a confession: I’m actually a computer program. And not even a very sophisticated one at that. Whew. Feels good to get that off my virtual chest.

[photo credit and quoted story credit: Northwestern University, Intelligent Informations Lab]