The word Skitag means “ski day” in German. The spirit of Skitag is to press the start button, record all your ski downhill throughout your ski day, and don’t think about your device until the day is over. Then take a look at what you have done. Learn from your metrics. Have a better understanding of your ski and carving style. Share your thoughts with your families and friends.Skitag AI Ethics and AI System lifecycle
Because it takes advantage of the embedded smartphone's accelerometer and gyroscope to learn your CARVING BIOMETRICS (your CARBIO). Because its AI System can identify your ski runs as well as your ski turns. Skitag can isolate each one of your turns, measure some relevant metrics (e.g. speed, style and path), and show you that information. You can make it an actionable information, and improve your technique. Or you can just have a better self-awareness of your ski moves.
It’s a machine learning algorithm that receives an input (e.g. speed, altitude, acceleration, attitude, etc.), process that Data, and outputs a binary decision (e.g. carving, not carving). The machine learning algorithm is a Long Short Term Memory (LSTM) Recurrent Neural Network architecture. It is used to evaluate the kind of information that has a timeline dependency.
Step 1: Download the Skitag App from the App Store. No user nor password needed.
Step 2: Go to your favorite Ski Resort. Get ready to start enjoying an augmented ski experience. Open the Skitag App. Press the start button. A stopwatch will start to count your ski session timing. It means the device is now recording. Wear your mobile in a place that is comfortable and stable (e.g. in your trousers pocket). Using mobile protection cases might keep the device more stable and protected in case of crash.
Step 3: Record your first FOUR downhills. It is the minimum quantity of information that the Skitag AI System needs to build your customized AI engine.
Step 4: Go to the DOWNHILLS screen. Select one DOWNHILL. Press the WAVEFORM button to access the labeling tool. Tap the button in the center of the screen to navigate between the three possible ski paths: Pitch, Roll and Yaw. Press the WAVEFORM+ button to save that choice. Once you select a path (e.g. yaw), the rest of the downhills will only show you that path. If you want to change it, you must delete your CARBIO and start all over again. Note: There is no gold standard for all. It's 100% a custom decision. It will depend on how you are wearing your smartphone while you are skiing or snowboarding.
Choose the area that best resembles the kind of CARVING BIOMETRICS you are willing to measure; i.e. clean and clear turn paths. This action will amplify the area of interest so you can have a better look at your turns details and shape. Choose the best points that resemble your CARVING. For a turn, whether it is a right or a left turn, choose only the points that are far away from the center. For the transition path, choose the points that are located in the central area. Make sure you are choosing all possible points of your area of interest. Avoid labeling the outliers; i.e. noisy points located outside what looks like a regular path. Upload the labeled dataset by pressing the CHIP button. Repeat this process as many times as possible. Remember that you will need at least 4 downhills to train your first CARBIO. Afterwards, the next datasets uploaded will be used to update your existing CARBIO.
Step 5: Go back to the Home Screen. Press the CARBIO card. Press the PLUS button to train from scratch your first CARBIO, or the REFRESH button to update an existing CARBIO. Remember that you need at least four labeled downhills to complete this task. Wait a few seconds. It won't take long.
Step 6: On average you’ll see that your CARBIO’s accuracy is close to 70% or 80%. If you are lucky, or your carving data is perfect, you could see something close to 90%.
Step 7: Record another downhill. Repeat the labeling process of Step 4, and then repeat the training process of Step 5. This time it is an update of your existing AI engine.
Step 8: Repeat Step 5 until you feel comfortable with your CARBIO’s accuracy value. Remember, you need a CARBIO’s accuracy above 90% to start seeing your Ski Style, Ski Balance and Ski Path.
Step 9: Record your Ski Sessions. Remember to wear your iPhone always in the same place so your AI can accurately predict your downhill movements, and remember to make sure the stopwatch is running so the Skitag App is reading and processing your CARBIO.
Step 10: It is time to enjoy your Skitag Augmented Ski Experience. Take a look at your ski metrics while you are enjoying a break, while you are having your favorite drink together with your friends, or while you are going uphill in the chairlift. Just “Keep calm and Skitag everywhere!
Don’t worry, you will. As long as you carry your iPhone always in the same position (i.e. the screen facing your leg), and in the same place (i.e. your right pocket of your ski pants), the AI engine will learn to recognize your carving biometrics. Overall it takes one ski day (i.e. one “skitag”) to surpass that threshold. Have in mind that the AI engine will learn from your carving biometrics. You probably have an unnoticed pattern that the AI engine can easily find. Nevertheless, if you still are experiencing difficulties to get a good accuracy value, please e-mail us to firstname.lastname@example.org. We’ll take care of it.
The AI Engine is trained to classify your turns. Therefore, you’ll notice that the AI engine is not accurate if there is a huge gap between the count of your; i.e. you think that you have done 15 right turns but the App tells you that you just made 2. Nevertheless, the information like Carving Style, Carving Speed or Carving Path is not affected by the performance of your CARBIO.
Here is a checklist that we recommend you to try before going forward with other solutions:
A CARBIO can suffer from Data Drift. It happens when your ski style changes over time. If that change is huge, it is possible that your CARBIO experience difficulties to correctly classify your turns.
That's an easy one! Keep updating your CARBIO in a regular basis.
You can train a brand new CARBIO using brand new input data (i.e. new downhills). It takes 6/7 downhills to train a new CARBIO.
Please, contact us by e-mail to email@example.com. We’ll be very pleased to help you understand what is happening and to solve your problems.
It’s difficult to come up with a standard optimal ski curve definition. Most of skiers will agree that an optimal ski curve depends on the ski boards, on the track inclination, the ski velocity and many more conditions; i.e. equipment, snow, track, skier's weight.
Some suggest to pre-parametrise the algorithm with this kind of information; i.e. snow conditions and/or ski board specifications. Nevertheless, that decision would mean to move away from the essence of artificial intelligence; i.e. find patterns from raw complex nonlinear data.
In that sense we found that most skiers may also agree that an evidence of an optimal curve is when it feels that the ski board traces a perfect curve without drifting. It feels smooth, they say. Therefore we thought that an ellipse could be a good standard for an optimal curve path while you are skiing.
Skitag gets every ski curve optimal path replacing the ellipse formula parameters with the ski curve timestamp events and the maximum tilt during the ski curve. Respectively the semi-major and semi-minor axis depending on the shape of the curve (long or short curve).
Back in 2021 we were saying that: "Only the skiers experience with Skitag App, Skitag IMU device and the results of the skitagCI initiative will tell us if this approach is a good one or if we need some other kind of approach like offering an App with a training configuration so the skiers collectively trace and classify ski curves events."
And: "Yes! Certainly this could be a very interesting project for the future; i.e. let the anonymous interactions of skiers to draw the optimal ski or snowboard curve paths. Interesting..."
Now in 2023 we have introduced user customized AI on-device. Every skier will have the chance to teach its own AI system to learn from its carving biometric. Their experience will help us find if our hypothesis was right or not.
Surftag, is a spin-off project of Skitag inspired by our passion to sports. This App is also designed to test the algorithms, and improve the Skitag UX with Apple Watch.
First, you need to wear an Apple Watch. Turn your Surftag App and press the Start button to start recording your surf data. While you are surfing you will see a stopwatch, a waves and attempts counter, and the paddling and distance meters in the water. To stop recording swipe to the bottom of your Apple Watch screen and press the Stop button. A screen will show your Surftag Session summary. Then, open your iOS App and sync your Apple Watch data. Go to Activity Screen and choose your Surf Session. You'll see your waves plotted in a map and you'll also see your overall Surftag data.
A Surftag Session will show you the number of waves surfed, the average speed (km/h) while surfing a wave and the time and distance (meters) surfing a wave. You'll also have the total time and distance of your Surftag Session. The distance paddling and the distance and number of positioning manouvers to line up or to let a wave pass.
Please contact our Skitag Early Adopter center (firstname.lastname@example.org). Our customer support team located in Pamplona (Spain) will answer your question as soon as possible.