jankoc

@HCHolmberg1 @QualisysMocap The study of Pellegrini et al. is very interesting, but I can't see that it shows an efficiency decline of 8-15% over 90+ minutes?

@QualisysMocap 🙏❤️

#MilanoCortina2026 XC distance racing: Motion capture shows 8-15% efficiency decline over 90+ min. Winner isn't always fittest - it's who maintains technique quality under fatigue. Athletes with better technique retention finish higher. Read the study: sciencedirect.com/science/articl…

@skalskip92 Yes, definitely, this would be a great addition!

here's side by side comparison top: naive trajectories drifting with camera motion bottom: motion-compensated trajectories pinned to the field same tracking, same detections, just smarter coordinate handling

I'm experimenting with camera motion compensation for trajectory tracking when the camera pans, naive trajectories drift with the frame. With motion compensation, trajectories stay "pinned" to the world should I include it in trackers-2.2.0 we release this week?

@mfranz_on @crcouve Idea is brilliant, but it looks a bit strange that the curling stone starts accelerating again after the speed has stabilized, can't see how that can be possible? I guess it is the camera movement calibration that is tricky to get accurate enough...

@crcouve Thanks! It looks me a while to be honest because the camera angle change in time and this make things more complex but we have a v1 now, happy for that

Curling stone tracking v1 Olympic Games edition. How I made this first version: - YOLO for real-time stone detection - ByteTrack (Supervision) for consistent object tracking - SciPy for trajectory smoothing - OpenCV + NumPy for annotation, velocity, direction and acceleration visualization The colors indicate whether the stone is : accelerating → green stable → orange decelerating → red

@mfranz_on Yes, I have looked into correcting for that as well, but more work needed. @mdunnhofer has done some nice work where they correct for camera motion by image matching in alpine skiing x.com/i/status/20187…

@jankoc one problem it is also the camera movement which creates a loss of coherence in the coordinate system

Winter Olympics Games this year will be computer vision enhanced. Ski jumping has always fascinated me and understanding how the athlete's poses influence the jump is a very interesting metric. I used a simple Python script with Mediapipe and basic algebra for angles.

@mfranz_on ... although the accuracy in the metrics unfortunately does not seem to be good enough with one camera in the field for meaningful variables, but I'll keep searching:)

@jankoc Yep, you can obtain a lot of metrics by analyzing athletes' poses.

@skalskip92 What about taking a look a sprint cross country skiing? A very challenging case, especially to automatically re-identify athletes across scene cuts. Here a short example with manual re-identification, where SAM3 is used at the start of each scene.

winter olympics are coming. time to have some fun. RF-DETR + Trackers hockey demo and more coming soon: github.com/roboflow/sports

Thanks, lots of great new stuff in there! Captions are great, new pointer is great, text overlay is great. Looking forward to some small improvements to the captions in the future: Manual captions work very well after you first have generated automatic captions using "Generate captions", but it would be great to be able to add manual captions also without using the auto-captions first (for my language the auto captions don't work well, and also in some cases I'd just want to make a few manual captions anyway). Also auto-sorting the captions on time in the video would be good

It’s finally live. I can get a full night’s sleep.

@JornSundby @PetterSkinstad Hva med å legge til noen bonussekunder til topp 3 i tillegg, så blir det enda litt mer utfordrende taktisk...?

@PetterSkinstad Synes konseptet fungerer bra som en del av en Tour. Det var mye bedre enn fryktet og tror det kan fenge litt utover kjernepublikummet til langrenn. Men TV-grafikken er jo helt krise da. Som vanlig. FIS har en stor oppgave med tv-produktet langrenn.

No' så dumt, men også noe bra og ikke minst et enormt forbedringspotensialt. Hva synes du om denne øvelsen? tv2.no/sport/no-sa-du…

Tror ikke det hadde hjulpet å velge andre TV-passeringer, her mangler forståelsen, og det så ikke ut som om dette var noe de har øvd seg på heller. Finnes det ikke løsninger som gir forslag til valg av kamera/scenevalg basert på mellomtider, @JornSundby ? Det burde jo være rimelig enkelt å lage et automatisert system som fungerer mye bedre enn den sveitsiske produsenten?

Ser 10 skate frå Davos i opptak. Evalueringa frå i fjor hjå produsenten ser ikkje ut til å ha inneholdt kva passeringspunkt dei skulle ha neste år for å bli betre. Her ville 2,7 , 6,7 og mål funka fint. Når skal langrenn få produsent som sikrer kvaliteten på intervallstart?

Check out the 62 nominations for the "Orienteering Course of the Year" in this ScrollOstory! Very cool to scroll through all the great courses, read the nomination texts, zoom in on the interesting legs and think through how to best solve the legs. Thanks to everybody for their nominations, and of course also to the sponsors! Check out the ScrollOstory and then head over to vote and win prizes worth more than 7000 Euro! omaps.worldofo.com/modern/lib/cou…

Sure, but if the morning reading is slightly low and you have a very good trend throughout the night, then that is a very good sign. If the morning reading is slightly low and you had a very poor trend throughout the night, that's another story. I definitely agree with you, but this added info is surely useful...

@hjluks The development of HRV through the night often gives very good insights, especially useful if you get a higher or lower reading than expected for the night and/or morning reading.

Tusen takk! Jeg er mest interessert i frem i tid, og det får jeg vel ikke ut av klimadatabasen. Så såvidt jeg kan se må jeg nok hente noen tusentalls lat/lng-punkter fra yr sitt API for å få ut data til å lage slike kart såvidt jeg kan se, litt avhengig av hvor stort område og oppløsning jeg ønsker.

@jankoc @jankoc Vi har ikke en karttjenste for det du etterlyser. I vår klimadatabase (seklima.met.no) kan du velge parameter for valgte perioder.

Må paraplyen frem denne uken🌂☔? På kartet ser du hvor det vil komme mer enn 10 mm nedbør mellom onsdag og fredag. Øst- og Sørlendingene vil nyte sol og gode temperaturer 🌤️🌡️ Troms og Finnmark vil også slippe unna mesteparten ⛅🍂 Resten av landet må derimot finne frem regntøy!

Late night work doing the in-depth analysis of the #EOC2025 Sprint yesterday, but definitely worth it. This was not your regular, boring, flat city sprint - check it out!

@forskningno Interessant at redusert aktivitetsnivå ikke er nevnt i artikkelen i det hele tatt...

Verken gener, tomme karbohydrater eller kjemikalier kan alene forklare den globale fedmeepidemien, ifølge forskerne. forskning.no/fedme-miljo-ov…

Full article here - including a look at the second last control that caused a lot of disqualifications, even the first team finishing the course news.worldofo.com/2025/08/28/eoc…

Interesting courses in the European Orienteering Championships yesterday - but not easy to understand during the broadcast (or even afterwards based on the GPS-tracking) what actually happened in the most tricky parts - I took a look at possible route choices at some of the most interesting legs at World of O #EOC2025

@ammattipyoraily Impressive, great to see that the estimates are that close!

🇫🇷 #TDF2025 ⛰️⚡️ Tobias Johannessen (6th in GC) on major climbs and comparing power meter data to est. power. Onley (4th) and Vauquelin (7th) posted Strava files with power as well, but Johannessen was the one who didn't have safety function on. Files were until finish line.