Swing Vision AI line calls and Analysis

hamsterhuey said:
Thanks for the reply. Actually it is on a swingstick and it does have the full-court in view. I cropped it out just to keep the overlay with the names out so I didn't post them up online. In all of these rallies, the camera was completely static. I do work in the field of computer vision and it is surprising to me how poorly it is working in some of these situations and I can't tell if this is a bug of some sort impacting me or my device specifically. It also looks like they don't utilize sound cues at all to refine bounce timing which is surprising.
Click to expand...

the swingstick or similar makes the angle tilted which bothers me.
i need the horizon to be straight horizontal
 
hamsterhuey said:
Does everyone else have SwingVision make egregiously bad line-calls or bounce location estimates fairly often, even for very easy balls that it should get right? I haven't been all that impressed with the quality of line-calling and I find it hard to believe that this is meeting metrics for automated line-calling. At the moment it seems way worse than well-intentioned players calling balls in/out from my experience. For the record, I am using a swing-stick with the top extended all the way for the highest vantage point and an iPhone SE 3rd gen.

Some examples attached
https://imgur.com/a/gZt5eq3
https://imgur.com/a/v8oCG35
Click to expand...
Fellow swing stick user here. Honestly I barely use the line-calling feature. I mainly use Swing Vision for reviewing videos for improvement, and the way it cuts out dead time is really valuable, in addition to the filters. I've found if I pay much attention to the line calling there are often big errors, but the majority of the time the ball is relatively close, probably within a foot. But 5% errors over 600 shots is 30 shots which is pretty significant.
 
mptennis said:
Fellow swing stick user here. Honestly I barely use the line-calling feature. I mainly use Swing Vision for reviewing videos for improvement, and the way it cuts out dead time is really valuable, in addition to the filters. I've found if I pay much attention to the line calling there are often big errors, but the majority of the time the ball is relatively close, probably within a foot. But 5% errors over 600 shots is 30 shots which is pretty significant.
Click to expand...
Thanks. Yeah, I have typically overlooked it but last night's game had a lot of egregious misses to the point that it was significantly impacting the autoscoring... which was annoying as I wanted to share with my partners. I literally had to manually edit the winner of each point in our 3rd set tie-breaker to have it work. I'm usually fine with if the calls are reasonably close to where they are landing, but right now, like with the example above, they are often completely nonsensical at which point I can't trust any of the spray charts, and it even impacts the filtering in some cases when reviewing videos.

Agree that the biggest benefit is the auto-edited points, but even with that, sometimes the points are getting messed up when it makes egregious errors with the line-calling.
 
I feel like when auto-scoring made the full release, shot placement and line-calling accuracy dropped. It seems like there’s more instances of shot placements that are way off and when you edit the shot, typically the bounce timing is way off, throwing the placement and shot speed. In the past, line calling on the near side was hardly an issue, but now you get a minority of cases where the challenged call is wrong when you review the footage after (via the Apple Watch it was difficult to tell). And these cases don’t have obstructed views of the ball bounce or lines.

I’ve noticed that run around forehands almost always get classified as backhands now, when I didn’t notice it happening that frequently in the past too.

Consequently, I still prefer tracking points manually with the watch and have turned off shot placement, ball speed, shot type overlays when exporting since they are no longer accurate.
 
Thanks. Yeah I reprocessed my video again and the shot placement / line-calling accuracy seems to be quite atrocious. It would be one thing if it was missing things that were within half a feet or so, but right now, a lot of the bounces are being marked 4-6 feet away from where they land. As a new subscriber, this has been quite disappointing. There is obviously still value in an edited down video, but right now, even there, I am seeing points not being split up correctly because the shot placement is so poor at times that their algorithms have no clue what the correct point duration was and how to appropriately clip the video for the point.
 
AndrewUtz said:
anyone know anything about the new app Play/Replay? not available in my country but it appears to do similar stuff to swingvision
Click to expand...

It looks like it's a different thing. Complete infancy, it's YT channel has only 1 subscriber, requires external sensors that are not your phone and do not film the session, is only a tool to provide stats, line calling and analysis. IMO the reason most people use swing vision is to have a better tool to actually see yourself playing, with the automated cuts. Swing Vision's stats and line calling are completely messed up - at least in my case, and that may be related to camera positioning, but that is not the reason why I use the app.
 
swupnil said:
@AndrewUtz the better line calling requires 4K video, which costs us 4x as much to upload & store in the cloud. Thus we have to charge for it.

If we can someday figure out a way to record at 4K but save the video at 1080p, we’ll gladly bring the better accuracy to Pro. We’re not trying to gatekeep better accuracy at all - just trying to be reasonable about the increased COGS on our end.
Click to expand...
some random thoughts (regarding 4k as a means to improve accuracy)..
  • to reduce costs, can the localApp be smart enough to only upload snippets (around the lines?)
  • can their be tiered pricing that enabled 4k mode to be uploaded/processed
random thoughts regarding line calling accuracy
  • would be cool if there was a way to have 2-4 phones around the court sync'd together to form a more complete picture (and would motivate more folks to buy a subscription... presuming it's possilbe, and line calling was accurate, imagine if every tournament required competitors to have a subscription and carry around a tripod/fence mount....)
  • what if there was a mode to only track a single line (if it made the line calling algo easier)... imagine having 6 phones setup (2 baseline, 2 sidelines, 2 svc lines)... i'm sure tennis parents would buy it (so could probably have at least 4 per match)
 
@swupnil Did the User interface change (for the worse) on Apple Watch ?

I had been using the Apple Watch for scoring matches but the score no longer shows up on Point+ mode the last couple of months. On version 11.6 (39.22)

Also, would love to be able to have more than just the "rally" and "Match" options show up on the Apple Watch ("Serve practice" is available only on the Iphone app but not on the Apple Watch? )
 
@hamsterhuey the screenshots you shared are quite surprising. We’ve been rigorously testing with USTA and the line calls (and shot placements) have been nearly perfect, certainly well better than a human.

In general, the app is optimized to accurately estimate “in” or “out” as accurately as possible. The actual 3D location within each region of the court is less accurate, but is significantly better in our testing compared to what you shared.

My only guess is that those additional short court lines on the court are messing things up and have low representation in our training set. USTA junior tournaments are never played on such courts and most of our customer base does not play on such courts.

Probably a data issue but thanks for reporting! If you can DM me a link to your session, we’d love to get it in the training set so that accuracy can improve.
 
Last edited:
@forzmr_b our accuracy metrics for in/out and 3d location are the highest they’ve ever been across a variety of courts where we’ve set up a 16 camera system to validate the accuracy. It would be super helpful to get session links from you so we can take a closer look (feel free to DM).

We’ve just come off a few junior tournaments with USTA where SwingVision provided line calling on 10+ courts and the accuracy was really solid.

I’m honestly very surprised to hear your feedback but would really like to look into this.
 
@nyta2 the line calling is all done on device so there’s no need to only “upload” content near the lines. The solution is basically what I mentioned before: process 4K but save in HD. This will likely only be possible on the very newest phones since downsizing every frame from 4K to HD in real-time at 60fps is super intensive.

We don’t have plans to support multiple cameras around the court or have cameras on specific lines. Over the last 5+ years we have seen other competitors do this. They have all come and gone. It’s just way too much friction and the market for that is just a very tiny handful of super users. It simply doesn’t scale and will never be a sustainable business model.
 
swupnil said:
@forzmr_b our accuracy metrics for in/out and 3d location are the highest they’ve ever been across a variety of courts where we’ve set up a 16 camera system to validate the accuracy. It would be super helpful to get session links from you so we can take a closer look (feel free to DM).

We’ve just come off a few junior tournaments with USTA where SwingVision provided line calling on 10+ courts and the accuracy was really solid.

I’m honestly very surprised to hear your feedback but would really like to look into this.
Click to expand...
My experience as I have mentioned earlier in this thread is similar to @forzmr_b which is to say, in/out accuracy is not great. This is why I suggested adding an option to disable auto scoring, because the inaccuracy sometimes leads to scoring inconsistencies.

I think it’s frankly more surprising that you are surprised, as several other people in this thread (not just myself and @forzmr_b ) have commented that line calling is unreliable, and that seems to be the general feedback on Reddit, as well, at least from what I have seen.
 
@thecatch33 yea I guess there is a lot of variability out there that we haven't accounting for. hamsterhuey's videos have those other court lines on top while forzmr_b's videos are zoomed out, have curvature on the baseline (usually due to third-party wide angle lenses) and the session says "Suboptimal Setup" within the app.

If you're not seeing the Suboptimal Setup and still finding clear in/out issues, then yes I am very surprised. We have thousands of support tickets and any time there is inaccuracy, it's almost always that. When people setup as intended, the results have been really good. Hard to standardize for such a consumer product but we're doing our best to improve everywhere we can.

As to your Auto Scoring concerns, there is supposed to be a Revert Scoring button if you don't like the results. I will double check.
 
swupnil said:
@thecatch33 yea I guess there is a lot of variability out there that we haven't accounting for. hamsterhuey's videos have those other court lines on top while forzmr_b's videos are zoomed out, have curvature on the baseline (usually due to third-party wide angle lenses) and the session says "Suboptimal Setup" within the app.

If you're not seeing the Suboptimal Setup and still finding clear in/out issues, then yes I am very surprised. We have thousands of support tickets and any time there is inaccuracy, it's almost always that. When people setup as intended, the results have been really good. Hard to standardize for such a consumer product but we're doing our best to improve everywhere we can.

As to your Auto Scoring concerns, there is supposed to be a Revert Scoring button if you don't like the results. I will double check.
Click to expand...
@swupnil It isn’t easy to achieve the optimal setup. Which would be better, zooming out .8x without wide angle or zooming in 1.1x with wide angle?
 
@forzmr_b the best is to use the standard 1x with no third-party lens if that allows you to see all 4 corners of the singles court and even just a few pixels beyond the corners.

If you’re unable to see the court, I would first recommend 0.9x (or 0.8x zoom if absolutely necessary) if you have the newer ultra wide lens introduced in iPhone 13 pro / 14.

If your phone is older then that, then last resort is a Xenvo lens.
 
swupnil said:
@forzmr_b the best is to use the standard 1x with no third-party lens if that allows you to see all 4 corners of the singles court and even just a few pixels beyond the corners.

If you’re unable to see the court, I would first recommend 0.9x (or 0.8x zoom if absolutely necessary) if you have the newer ultra wide lens introduced in iPhone 13 pro / 14.

If your phone is older then that, then last resort is a Xenvo lens.
Click to expand...
I use an iPhone 13 mini so no in-built wide-angle lens. In the past, I had to zoom out to capture both singles sidelines. I’ll see if I can see the lines with 0.9x. If I can’t, would 0.8x w/o Xenvo be better or 1.1x with Xenvo? Whichever gets me closer “crop” of the 4 corners?
 
forzmr_b said:
I use an iPhone 13 mini so no in-built wide-angle lens. In the past, I had to zoom out to capture both singles sidelines. I’ll see if I can see the lines with 0.9x. If I can’t, would 0.8x w/o Xenvo be better or 1.1x with Xenvo? Whichever gets me closer “crop” of the 4 corners?
Click to expand...
Quick update. Without the Xenvo I have to rely on 0.8x zoom and results weren’t great. I also realised the next increment for zooming in was 1.5x so I didn’t bother trying to zoom in with the Xenvo attached. Guess I’ll have to make do with my Xenvo at 1x and accept the sub-optimal setup.
 
forzmr_b said:
Quick update. Without the Xenvo I have to rely on 0.8x zoom and results weren’t great. I also realised the next increment for zooming in was 1.5x so I didn’t bother trying to zoom in with the Xenvo attached. Guess I’ll have to make do with my Xenvo at 1x and accept the sub-optimal setup.
Click to expand...
I'm on a 13 pro and for doubles on normal 1.0x i still get a suboptimal setup because it doesn't get the near side corners. Do you think 0.8 would make it optimal?
 
@forzmr_b yea better to use 1x with Xenvo at that point I feel. The wide lens is more dependable than the ultra wide so any zoom level about 1 should give better results.

@kevonian 0.8x should be okay but may suffer a bit in low light. Generally the right way to go for doubles though.

@mptennis best place to submit feature requests is here and/or support@swing.vision. Feature roadmaps a bit trickier but I want to find a way to do it more publicly at some point. Add missing points, we are discussing internally whether that is necessary or if we can find a way to resolve with auto scoring improvements.

In general, it is SUPER helpful if you can send match links of autoscoring issues to support@swing.vision - engineers are following this very closely now and trying to document this more thoroughly so we can have improvements in every release.
 
To all who use a wide angle lens attachment. The app now uses the other rear camera lens, at least for the 13 mini. I positioned it over the usual lens and inadvertently blocked the other and all I recorded was black! Bummmed because I hit some nice shots

@swupnil Would appreciate if users are informed about such changes in the future, thanks!
 
@forzmr_b we haven't changed anything as far as I know. If you select 1x or higher, we use the wide lens. If you use 0.9x or lower, we use the ultra wide lens. This should be the case for any phone. The only exception is the selfie camera on iPads that have Center Stage.
 
swupnil said:
@forzmr_b we haven't changed anything as far as I know. If you select 1x or higher, we use the wide lens. If you use 0.9x or lower, we use the ultra wide lens. This should be the case for any phone. The only exception is the selfie camera on iPads that have Center Stage.
Click to expand...
Apologies, my bad. I accidentally pressed the flip camera button so my wide angle was covering the front camera!
 
swupnil said:
@forzmr_b we haven't changed anything as far as I know. If you select 1x or higher, we use the wide lens. If you use 0.9x or lower, we use the ultra wide lens. This should be the case for any phone. The only exception is the selfie camera on iPads that have Center Stage.
Click to expand...
What is different with the selfie camera on iPads that have Center Stage?
 
Have the Swing vision servers been having problems recently? My teammates and I have all had issues with uploading and corrupted videos and being unable to watch videos over the last 3 days.
 
Demented said:
720p and 60 fps will yield better results if you can't do 1080 and 60. The frames are more important than the clarity. The height of the camera is also very important. I find that 10 feet is really the minimum in order to get the net low enough that it's below the service line. 12 feet is better. Also every shot is going to be different due to factors such as players or the net occluding the bounce. When the software can clearly see the hit and the bounce, it's fairly consistent. Other systems such as playsight typically rely on 4 or more cameras to ensure that the ball is always visible. With just 1 camera, you're always going to be at the mercy of visibility.
Click to expand...

so I was one of those people that tended to do 1080p at 30fps because whatever I don't care about the ball speed or placement so much as the other stuff.

only the latest version of the app will no longer allow any imports unless the video is at least 720p at 60fps. if that will provide a better experience, great - but now I have videos I can't process because there was no warning of this change beyond the app updating, trying to use it like I always have, SURPRISE your previous workflow no longer works.

this should have been socialized much, much better. and near as I can tell their docs/website talk about how video has a max of 1080p/60fps - but there is no published min until the new import error dialog. if you can introduce a nanny dialog at the import step you could have for sure had a warning dialog like WARNING 1080p/30fps IS GOING AWAY vs. the HAHAHA NO MORE IMPORTS FOR YOU deal we're getting now, instead.
 
@ohplease ah sorry about that! We totally should’ve given a heads up. 30fps imports are like less than 0.1% of sessions so we totally missed this. If you DM me your Apple ID email address, we’d be happy to add you to TestFlight so you can download an older version of the app. Apologies again
 
How do we clear the app data. I’ve deleted the video from the me section after exporting it to my photos but in my ipad storage it still shoes swingvision is holding on to a ton of storage and data.
 
I think it might be a processor problem. My M1 iPad Pro still doesn't work with swing vision since the iOS update. But I tried my A14 iPad 10 and it works normally with swing vision.
 
@swupnil - Is there a way to review the serve practice footage after a few serves without taking down my phone from the fence? Like could I hit a few serves, then use an iPad to watch the video to check my technique w/o stopping the recording on my phone? It seems like you can do that already with challenging line calls, but wasn't sure if you could do that with serve practice mode.
 
Is there a way to clear the temp storage in this app. I have it edit my videos and then delete it in the app after I’ve exported the video locally to my ipad photos app. So there’s no video in the swingvision app but it’s still taking up 20gb of space in the storage according to settings. Only way is to delete the app, redownload it and sign back in.
 
2ndServe said:
Is there a way to clear the temp storage in this app. I have it edit my videos and then delete it in the app after I’ve exported the video locally to my ipad photos app. So there’s no video in the swingvision app but it’s still taking up 20gb of space in the storage according to settings. Only way is to delete the app, redownload it and sign back in.
Click to expand...
Would like to know too. Currently, I try to offload sessions before deleting them. Seems to help. In the past, I just deleted them. My other hypothesis is failed imports/crashes clogging the temp folder. Likely only recourse is reinstalling the app.
 
Why I am not renewing my Swing Vision Account Pro after one year
Hi Swing Vision community, hi Swupnil,
here I would like to share my experiences after one year of using the Swing Vision App + Stick and explain why I no longer use the Swing Vision App (although I would actually “need” it if it would deliver what it at least suggests to promise:

1. accuracy (line calling)
2. scoring
2. analysis after the game
3. handling swing stick

My intention is to help SwingVision and steer the development of the app in a more useful direction Of course, I would also be happy to receive feedback from other users as to how my experiences match theirs or whether I have not yet used (understood) some functions in the app correctly.

The line calling, the precise tracking of all game-relevant information, the scoring of a game are all functions that I would very much appreciate in tennis and would also be prepared to pay around 150 - 170$ [€] for.

I was very enthusiastic about the concept at the beginning and was even willing to switch from Android to Apple just for Swing Vision (I haven't regretted this switch so far ;-))
My experience started with the Swing Vision Pro app and with the iPhone 15 max + Apple Watch 9 here in Germany in the tennis hall on the carpet exactly one year ago.


1. accuracy (line calling)

I was initially surprised at how accurately the app calculated the balls in the court after my first quick tests.
Later, however, on closer analysis on the PC or in the app, I noticed that 5-10% of the balls in the field were calculated incorrectly. I became suspicious when we used the app's “line calling” function for a ball close to the line in a doubles match. I trusted the actual result of the app, but I saw that the line-calling function did not show the previous shot in the field, which clearly landed in the field. I then became more suspicious and started watching the recorded games on the PC to check the line calls and speeds. The result was somewhat disappointing= 4-8% error rate!
I could live with the result and told my playing partners that not all shots, especially those further away from the camera, were correctly recorded by the system. The clay court season starts here in May and I was looking forward to testing the Swing Vision app with even better (brighter) lighting conditions. I did serve practice and practiced with the ball machine on a clay court and had everything recorded and tracked with the Swing Vision app.
Even during the service training, I noticed that many balls that were out of bounds were considered to be in the court and vice versa (thanks to the app's acoustic exclamations from the Iphone, I was able to check this immediately - a really very good function). It was a similar picture during ball machine training, with the result that the error rate was significantly higher outside in the open air on a clay court.
Especially, of course, because the camera is behind you and has to calculate the balls coming up in the field further back. Of course, shadows from trees or poor line maintenance on the clay court could also have had an influence on the poor result. Unfortunately, I don't know of any courts for amateur players that have better conditions (less shade, better line maintenance). After the 3rd recording on a clay court, I gave up using Swing Stick because the results didn't add any value. I have now tried the app again in the fall under a bright cloudy sky. The results were still not good. Especially the baseline further away from the camera is apparently not correctly recognized (calculated) by the camera.

2. scoring
Scoring via Apple Watch works very well and is also a real added value. Autoscoring has never worked for me just by pressing a button. Apparently, individual calculated points in the app have to be edited again and again for the autoscoring to work. This is not a bad thing, as every AI system causes errors or provides misinterpretations in certain unforeseeable situations.
But - and here comes the big disadvantage of this app for me - unfortunately these points can only be edited on the iPhone and not in the browser on the Windows desktop. Searching through hundreds of shots in the app and editing them with the slider on the relatively small touch screen of the iPhone is very inconvenient. Here I would rather work with a PC mouse and use a larger screen. I have therefore given up this function (auto-scoring / shot post-processing) and then I come to the next point ...

3. analysis after the game
The post-match analysis is not very useful, as many balls are not displayed correctly or there are recorded shots that you don't want to include in your statistics. Here you would have to invest some time and effort in editing the shots afterwards. But that brings me back to the point I mentioned earlier: editing hundreds of clubs on an iPhone. No thanks!
There is also room for improvement with the existing statistical key figures. The model “Tennis Math” app for Android from “Seven Courts” (also available soon for IOS) has a lot more key figures to offer.
At the moment, for example, if I'm really interested in analyzing a game afterwards, I use the swing stick and simply record a game without a cut. I then analyze the game with the Tennis Math app. Also awkward ball by ball - unfortunately! But I can operate the video on the screen. For example, I can use slow motion frame by frame with any mpeg player on the PC. The tennis app “Tennis Maths” runs on the side on the smartphone. Here I enter the strokes point by point and have more precise and in-depth statistics calculated than with the Swing Vision app. Very time-consuming, admittedly! But unfortunately Swing Vision doesn't offer me any added value here. I get simpler and faster analyses, but they are inaccurate and don't go deep enough!

4. handling Swing Stick
Tennis is generally very hectic. Matches want to be started immediately, as the courts are not available indefinitely and the time for a long game is limited anyway. That's why installing the stick could be a little easier and quicker. Especially in the hall, where you usually only have a curtain to attach it to, the installation is a bit fiddly. The correct angle and tilt for the camera perspective is quickly shifted by moving the curtain. The installation time of 5 minutes quickly turns into 10 minutes. You don't get this time from your fellow players and certainly not if the results of the app are not accurate... :-( For these reasons alone, I have sometimes refrained from recording.

Conclusion:
The SwingVision app is a great concept, which I would be willing to pay money for. But the system is not yet fully developed.

Which I can't understand at all: @ Swupnil you wrote here in the forum about half a year ago that you wanted to include another camera recording in this concept. Now you write that you have no intention of doing so. I think the results would be many times more accurate if, for example, a second IPhone of a fellow player recorded a game so that both baselines and service lines could be recorded well. I find it disappointing to stop further development here.
I would gladly become a SwingVision customer again, especially if the accuracy is right.
Only when the accuracy is right ...
- the acceptance of using line calling among fellow players will increase again.
- the benefits of using the app in training will increase again.
- the analyses are meaningful
- it makes sense to invest time in the post-processing of analyses
If an application were then programmed that could be operated via a browser or a separate app on the PC, the SwingVision world would be almost perfect for me.
 
partycamel2002 said:
Why I am not renewing my Swing Vision Account Pro after one year
Hi Swing Vision community, hi Swupnil,
here I would like to share my experiences after one year of using the Swing Vision App + Stick and explain why I no longer use the Swing Vision App (although I would actually “need” it if it would deliver what it at least suggests to promise:

1. accuracy (line calling)
2. scoring
2. analysis after the game
3. handling swing stick

My intention is to help SwingVision and steer the development of the app in a more useful direction Of course, I would also be happy to receive feedback from other users as to how my experiences match theirs or whether I have not yet used (understood) some functions in the app correctly.

The line calling, the precise tracking of all game-relevant information, the scoring of a game are all functions that I would very much appreciate in tennis and would also be prepared to pay around 150 - 170$ [€] for.

I was very enthusiastic about the concept at the beginning and was even willing to switch from Android to Apple just for Swing Vision (I haven't regretted this switch so far ;-))
My experience started with the Swing Vision Pro app and with the iPhone 15 max + Apple Watch 9 here in Germany in the tennis hall on the carpet exactly one year ago.


1. accuracy (line calling)

I was initially surprised at how accurately the app calculated the balls in the court after my first quick tests.
Later, however, on closer analysis on the PC or in the app, I noticed that 5-10% of the balls in the field were calculated incorrectly. I became suspicious when we used the app's “line calling” function for a ball close to the line in a doubles match. I trusted the actual result of the app, but I saw that the line-calling function did not show the previous shot in the field, which clearly landed in the field. I then became more suspicious and started watching the recorded games on the PC to check the line calls and speeds. The result was somewhat disappointing= 4-8% error rate!
I could live with the result and told my playing partners that not all shots, especially those further away from the camera, were correctly recorded by the system. The clay court season starts here in May and I was looking forward to testing the Swing Vision app with even better (brighter) lighting conditions. I did serve practice and practiced with the ball machine on a clay court and had everything recorded and tracked with the Swing Vision app.
Even during the service training, I noticed that many balls that were out of bounds were considered to be in the court and vice versa (thanks to the app's acoustic exclamations from the Iphone, I was able to check this immediately - a really very good function). It was a similar picture during ball machine training, with the result that the error rate was significantly higher outside in the open air on a clay court.
Especially, of course, because the camera is behind you and has to calculate the balls coming up in the field further back. Of course, shadows from trees or poor line maintenance on the clay court could also have had an influence on the poor result. Unfortunately, I don't know of any courts for amateur players that have better conditions (less shade, better line maintenance). After the 3rd recording on a clay court, I gave up using Swing Stick because the results didn't add any value. I have now tried the app again in the fall under a bright cloudy sky. The results were still not good. Especially the baseline further away from the camera is apparently not correctly recognized (calculated) by the camera.

2. scoring
Scoring via Apple Watch works very well and is also a real added value. Autoscoring has never worked for me just by pressing a button. Apparently, individual calculated points in the app have to be edited again and again for the autoscoring to work. This is not a bad thing, as every AI system causes errors or provides misinterpretations in certain unforeseeable situations.
But - and here comes the big disadvantage of this app for me - unfortunately these points can only be edited on the iPhone and not in the browser on the Windows desktop. Searching through hundreds of shots in the app and editing them with the slider on the relatively small touch screen of the iPhone is very inconvenient. Here I would rather work with a PC mouse and use a larger screen. I have therefore given up this function (auto-scoring / shot post-processing) and then I come to the next point ...

3. analysis after the game
The post-match analysis is not very useful, as many balls are not displayed correctly or there are recorded shots that you don't want to include in your statistics. Here you would have to invest some time and effort in editing the shots afterwards. But that brings me back to the point I mentioned earlier: editing hundreds of clubs on an iPhone. No thanks!
There is also room for improvement with the existing statistical key figures. The model “Tennis Math” app for Android from “Seven Courts” (also available soon for IOS) has a lot more key figures to offer.
At the moment, for example, if I'm really interested in analyzing a game afterwards, I use the swing stick and simply record a game without a cut. I then analyze the game with the Tennis Math app. Also awkward ball by ball - unfortunately! But I can operate the video on the screen. For example, I can use slow motion frame by frame with any mpeg player on the PC. The tennis app “Tennis Maths” runs on the side on the smartphone. Here I enter the strokes point by point and have more precise and in-depth statistics calculated than with the Swing Vision app. Very time-consuming, admittedly! But unfortunately Swing Vision doesn't offer me any added value here. I get simpler and faster analyses, but they are inaccurate and don't go deep enough!

4. handling Swing Stick
Tennis is generally very hectic. Matches want to be started immediately, as the courts are not available indefinitely and the time for a long game is limited anyway. That's why installing the stick could be a little easier and quicker. Especially in the hall, where you usually only have a curtain to attach it to, the installation is a bit fiddly. The correct angle and tilt for the camera perspective is quickly shifted by moving the curtain. The installation time of 5 minutes quickly turns into 10 minutes. You don't get this time from your fellow players and certainly not if the results of the app are not accurate... :-( For these reasons alone, I have sometimes refrained from recording.

Conclusion:
The SwingVision app is a great concept, which I would be willing to pay money for. But the system is not yet fully developed.

Which I can't understand at all: @ Swupnil you wrote here in the forum about half a year ago that you wanted to include another camera recording in this concept. Now you write that you have no intention of doing so. I think the results would be many times more accurate if, for example, a second IPhone of a fellow player recorded a game so that both baselines and service lines could be recorded well. I find it disappointing to stop further development here.
I would gladly become a SwingVision customer again, especially if the accuracy is right.
Only when the accuracy is right ...
- the acceptance of using line calling among fellow players will increase again.
- the benefits of using the app in training will increase again.
- the analyses are meaningful
- it makes sense to invest time in the post-processing of analyses
If an application were then programmed that could be operated via a browser or a separate app on the PC, the SwingVision world would be almost perfect for me.
Click to expand...

Excellent analysis, agree with pretty much all you say. SW just isn't worth it at the moment. Maybe with multiple cameras in future? Certainly it won't take off in my neighbourhood until they have an Android app too...
 
You must log in or register to reply here.

Similar threads

D
problems with Swing Vision video playback
Replies
3
Views
1K
waterpro
W
A
Swing Vision: great app, appalling customer support
Replies
3
Views
803
DrTompu
DrTompu
ruhil
Using Maps and Data Visualization to Understand Tennis
Replies
0
Views
601
ruhil
ruhil
N
how to best measure serve speed, and compare to pros...
Replies
28
Views
8K
Injured Again
I
S
Electronic Line Calls
Replies
2
Views
899
West Coast Ace
West Coast Ace
Back
Top