! This post is also available in the following language. Japanese

[Event Report] LINE Fukuoka Developers Host Hackathon

Held for the first time in 2016, the immensely popular Hackathon—held by the engineers in LINE Fukuoka’s Development Department— recently came to a close. Just like its inaugural event, many of the engineers enthusiastically took part in this occasion. They were given a time frame of two days each to develop and present their ideas, and the final demonstration took place in Beppu of Oita Prefecture, which is well-known for its Jigoku Meguri (Hell Tour) and onsens (hot springs). The engineers achieved the feat of transforming all 16 ideas into prototypes despite the tight two-day development schedule. Nagashima-san, an intern for the LINE Fukuoka Data Analytics Team, brings us this report on the Hackathon.

The majority of demonstrations revolved specifically around new technologies: drone programming, AR/VR, Blockchain, big data analysis, and image recognition. It was a potent reminder of the engineers’ insatiable sense of curiosity for anything cutting-edge, and their capacity to efficiently implement new technology. Also, every concept was not just technologically stirring, but simply enjoyable to watch in action, which was clear from the constant laughter from the audience.

There were two demonstrations that were especially memorable. One was a glove called “Gocky”, which can play various musical instruments when the user adjusts the angle of their fingers. Another was the high-speed “touch machine” that automatically hammered out super high scores in one of LINE’s games. The “touch machine” can recognize the game screen (image recognition) of Jungle Pang, which is one of LINE’s games.

Hackathon really is the greatest event for engineers. They have a fun time developing while being given the chance to interact with technologies they’ve never seen before and actually learn a lot.

Yogo-san looking giddy while trying on the original glove invention “Gocky,” and making the final adjustments before the presentation.

As much as I would’ve liked to have written about all 16 ideas, I’m just going to describe the three that were especially popular.

1. LINE QUEST / Dmitry Ratushnyy, Yusuke Onishi, Mathiaszime Aboudoubagassi, and Makoto Umami

Simply put, this service incorporates game elements (the user needs to clear quest challenges) and Augmented Reality (AR) into a conventional map application. When users upload images of locations they have visited, their pictures appear as if suspended in the real world when other users visit the same place on the map application. The “quest” part is completed once the user arrives at the location on the map where the red pin is, at which point it turns green. In addition, users can receive kudos from LINE characters depending on how long it takes them to clear a pin.

During development, ARkit and Scenekit were used for AR implementation. Furthermore, to grasp the precise location of users who were indoors, engineers used beacons to increase the accuracy of measurements. The service also comes with a built-in mechanism that prevents users from cheating, since the codes are unique only to specific locations. This leaves the user with no choice but to walk to the destination.

2. Sharing Smile Drone / Takeo Nanba, Hidenori Takeshita, and Kunimi Nagakura

This idea combined two groups that were involved in drone programming during the Hackathon. The team led by Namba-san, Takeshita-san, and Nagakura-san developed a hands-free photography feature by controlling a drone. Utilizing the voice recognition function in Apple Watch, the user would speak a command, which is then processed via iPhone. The user can order the drone to take off, rotate, and land; for a group photo, the drone autonomously finds the coordinates where all members will fit into the camera frame. Once in position and after detecting everyone’s smile through facial recognition technology, the drone takes a photo which is then sent to LINE Notify, allowing users to edit it through SNOW later on.

The drone model, Parrot Mambo FPV, was used in the demonstration, and for drone control, Parrot SDK was adopted in addition to Vision.Framework, which was employed as the library for facial recognition. Since the Parrot SDK sample covers the basic scope of the drone’s attitude control, it makes programming development easy, even for beginners.

Some of the challenges the members faced during development were that they had to compress images taken by the FPV camera through the H.264 video-compression standard, which was followed by the difficult task of having to decode it to the color data of RGB. Furthermore, they had to repeatedly perform debugging operations, despite the fact that the drone’s battery lasted for only five minutes. Since the drone and the camera became overheated from continuous usage, they also had to reboot over and over after finding a way to cool them down as efficiently as possible. In the video, we can see the drone correctly executing orders and finding the spot to take the best smiling picture of the three engineers.

I thought the time lag it took for the drone to take off from the moment the command was given added to its charm.

3. Globe Infographics / Koji Lin, Chen Yucheng, Tsai Fengyi, Kengo Tateishi, and Atsushi Wada

Visually speaking, this was probably the most stunning demonstration of all. Infographics beautifully visualized LINE’s user data and activity logs onto a 3D world map that resembled a satellite photo. While the program is capable of visualizing various content, on the day of the presentation, the team members demonstrated in animation the “sending and receiving” of LINE stickers on the map between engineers living in different locations.

As it becomes evident from reading the presentation materials, this project really symbolized the moment when front-end engineering enjoyed the spotlight. This is because the development required a fair amount of UI technology, such as expressing the smooth connection of two points on the globe with the Bezier curve, and executing minute calculations as well as adjustments in order to properly display country names on the map, regardless of the view angle. In day-to-day analysis work which demands quick output, I’ve never come across such refined infographics. After this demonstration, I couldn’t help but feel inspired to learn more about data visualization technology that instantly captivates viewers.

How was the report? I didn’t get into details, but personally speaking, I really enjoyed the Jungle Pang demonstration that I mentioned in the beginning.

Even though one error was made, I thought it was sufficiently accurate. Apparently, the engineers utilized Convolutional Neural Network (CNN) during the initial stages of development in order to improve accuracy. If CNN is implemented in the future, a full combination of error-free maneuvers may not be a fairy tale. I got the distinct feeling that these button mashing games will be hacked through image recognition and machine learning in the future.

We took a group photo after the presentations. I don’t think it would have been possible to capture a better photo (even with the Sharing Smile Drone!).

Before and After Hackathon

After the Hackathon, it was time for hot springs and sightseeing! Please enjoy some of the candid photos of LINE Fukuoka engineers.

The atmosphere on the bus en-route to Beppu reminded me of a school field trip!

Looking great in their yukata robes while enjoying Japanese sake.