In/Out Device

I guess the app only check once a day for updates and you cannot force a check. It is now updated.

Is anyone still using this? Is the accuracy better? I fear that match mode has been delayed because it would be a mess if not accurate. Each bad call would need to be corrected at the machine each time which will hurt the flow of a match.

I have not just hit the ball lately with the machine so have not been paying attention to acuracy. I have been plagued with freezes, corrupted videos and stats that never synch back to my phone. Even an uninstall, shutdown and reinstall on my phone does not clear my app off completely, which is odd.
 
Great job by those of you who are testing the device.

Too bad it’s turned out how it is. One person just can’t do it all. And if he went with better cameras, it would probably move the price point too high.

Haven’t read every page - I assume shadows on a court would completely throw it off, correct?
 
I've been having hardware issues from the beginning. The original device he sent me had camera issues, and he finally replaced it. Now, the replacement device won't accept the batteries anymore. Plus, the last couple months, the batteries never show more than 50% capacity. Opened yet another ticket, but he doesn't respond. Good luck if you buy one. I'll use small claims court if this drags out too long.
 
Looks like the inventor is still going full ahead with updates and improvements. Updated website at http://inout.tennis/en/index.htm. There's a smartwatch companion app, re-enforced mount/cage to protect the device, updated housing, and new software versions for line accuracy coming. IMO this project will get there with enough time, it was just a very early launch (alpha/prototype).
 
He really needs to work on customer service first. I had my first device replaced because of defective cameras. The replacement device never had working audio, and the batteries started to only show 50% quickly - they could never record an entire match. Additionally, the device never did call lines properly. Eventually, the battery insert broke and won't accept the batteries anymore, rendering it useless. I filed *another* support ticket and waited and waited. Eventually, after complaining on his new YouTube feed, he replied and told me he won't warranty the device anymore - I cost him too much. He doesn't realize that shipping out defective units is costing himself, it is not my fault. Anyway, he made it clear that my business means nothing to him -- I see on his Twitter feed other users are suffering similar fates.
 
He really needs to work on customer service first. I had my first device replaced because of defective cameras. The replacement device never had working audio, and the batteries started to only show 50% quickly - they could never record an entire match. Additionally, the device never did call lines properly. Eventually, the battery insert broke and won't accept the batteries anymore, rendering it useless. I filed *another* support ticket and waited and waited. Eventually, after complaining on his new YouTube feed, he replied and told me he won't warranty the device anymore - I cost him too much. He doesn't realize that shipping out defective units is costing himself, it is not my fault. Anyway, he made it clear that my business means nothing to him -- I see on his Twitter feed other users are suffering similar fates.
Early adopter woes. Consider this a lesson learned. You are essentially an alpha tester that is helping the inventor out by finding flaws. You are not a consumer at this stage. You either take on an altruistic attitude that you are contributing to the development of a worthwhile product and the money you've invested goes to a worthy cause, or you take the attitude of burned consumer who learns the meaning of caveat emptor when investing in the early development of devices.
The creator of the device is a solitary inventor with a good idea. He's not Sony or Apple. What customer service could you provide as a single person?
 
The creator of the device is a solitary inventor with a good idea. He's not Sony or Apple. What customer service could you provide as a single person?
The product idea the inventor came up with here is a good one. There is certainly a huge market for such a product if it works and meets customer expectations.

I don't understand why he just didn't develop a basic Proto-type that demonstrates the product's potential. Then sell it off, or license it to a large Tennis related company like one of the racquet manufacturers or other global sporting goods companies.

As we know, the primary reason Elon Musk developed Tesla Motor Vehicles was to demonstrate the value of his Tesla Batteries. He is playing a Long Game and it will probably pay off big time for him.

History is littered with Inventors who did not make a lot of money out of great ideas simply because they could not let their "children" go so that they could grow. Inventors are usually very good at "Inventing". They are often not so good at developing or running Businesses.
 
i'm totally agree with Dartagnan64, even if it's sad for carter, but it was "the risk", and i haven't get my "courage" to take this risk until now... but i'm still really interested by the product even if it still in development.
On the contrary, karma tennis, I think it's a chance for us, that Gentil do not sell the idea, because:
> first the price for sure WILL NOT BE 199$ AT ALL !!! (as there is no equivalent product around this price...playsight,mojjo, hawkeye...not accessible, so it will be easy to sell it at 1000$ and it will be reserved to pro only...)
> second the product will be sell without saying that is beta version (many tennis sensor software or hardware aren't really finish...)
> and third, it will be not sure that you will have a better service customer !!! (i have paid 100 $ the head sensor and i can say that i have no complete documentation at this time, only 9 FAQ, very much less information on evolution of software)
on the contrary,look all informations we can have with Gentil each month, the "alone" guy who make all from A to Z.

The unique risk is that he do not sell enough product during development, so that he will may be forced to stop all.
 
Early adopter woes. Consider this a lesson learned. You are essentially an alpha tester that is helping the inventor out by finding flaws. You are not a consumer at this stage. You either take on an altruistic attitude that you are contributing to the development of a worthwhile product and the money you've invested goes to a worthy cause, or you take the attitude of burned consumer who learns the meaning of caveat emptor when investing in the early development of devices.
The creator of the device is a solitary inventor with a good idea. He's not Sony or Apple. What customer service could you provide as a single person?
On the flip side, I live in a state that provides an implied warranty for anything that I purchase. Why? Because our government recognizes that when a consumer puts his trust into a seller and gives money for the promise of a working item, then that item needs to live up to the expectations that any reasonable person would have from it. While the software was labeled "beta", there have been promises of a working system in the next 6 months. I can wait for that. It is also implied that the product can live up to its specifications. This hasn't happened. I've twice received defective hardware, and I'd like the seller to fix or replace it to meet reasonable expectations.
If I get irritated enough, I have the right to take him to small claims court to recover damages. Additionally, a class-action lawsuit can be filed. But most reasonably, the seller's reputation is damaged by reports of these failures. This is not about me learning a lesson, but rather, the seller learning one.
 
The smart watch app is not available yet. I think this is really needed to execute on the final vision so I am happy it exists.It also seems to be that a second device will be required to use the doubles mode. That is disappointing but it makes sense because the second player will be in the way of the camera too often.

Waiting for the April update...
 
The product idea the inventor came up with here is a good one. There is certainly a huge market for such a product if it works and meets customer expectations.

I don't understand why he just didn't develop a basic Proto-type that demonstrates the product's potential. Then sell it off, or license it to a large Tennis related company like one of the racquet manufacturers or other global sporting goods companies.

As we know, the primary reason Elon Musk developed Tesla Motor Vehicles was to demonstrate the value of his Tesla Batteries. He is playing a Long Game and it will probably pay off big time for him.

History is littered with Inventors who did not make a lot of money out of great ideas simply because they could not let their "children" go so that they could grow. Inventors are usually very good at "Inventing". They are often not so good at developing or running Businesses.
I can see him not selling it. But should have taken on an investor (still retain the majority of the company for himself) so he could hire a few employees and delegate some of the work. Dealing with his supply chain by itself is a big job. Packaging, customer support would be easy to offload Too. He could stick to improving the software.
 
On the flip side, I live in a state that provides an implied warranty for anything that I purchase. Why? Because our government recognizes that when a consumer puts his trust into a seller and gives money for the promise of a working item, then that item needs to live up to the expectations that any reasonable person would have from it. While the software was labeled "beta", there have been promises of a working system in the next 6 months. I can wait for that. It is also implied that the product can live up to its specifications. This hasn't happened. I've twice received defective hardware, and I'd like the seller to fix or replace it to meet reasonable expectations.
If I get irritated enough, I have the right to take him to small claims court to recover damages. Additionally, a class-action lawsuit can be filed. But most reasonably, the seller's reputation is damaged by reports of these failures. This is not about me learning a lesson, but rather, the seller learning one.

The seller is just someone with a good idea that wanted to try to make a go of it and earn a living. Are you that upset about losing $200 dollars that you'd like to stifle this device entirely and put this man in the poor house with a class action suit? Just to teach him a lesson? That's pretty harsh.
I'm sure the seller has learned a lot of valuable lessons along the way. It's often a lot easier to invent something than to manufacture and market it. Hopefully some of these lessons will lead to better things for consumers.
 
I can see him not selling it. But should have taken on an investor (still retain the majority of the company for himself) so he could hire a few employees and delegate some of the work. Dealing with his supply chain by itself is a big job. Packaging, customer support would be easy to offload Too. He could stick to improving the software.
The idea is good, but I think either the inventor/creator had unrealistic expectations for the performance of his product and/or the customers did. Given the camera's limited viewing angles and performance and number of cameras, there was no way it was going to be that accurate -- especially at night.

He could have hired a few university physics students and they would have quickly told him the limitations of his idea and equipment. Some improvement can be made with software, but not a lot.

Training alone is not going to turn Woody Allen into elite an powerlifter.
 
@onehandbh i agree w your 1st paragraph. I saw the Bloomberg story - the inventor touted it as s game changer.

To the rest of your comment: no. Look him up. He doesn’t need any Physics students’ advice. He’s been very successful in his career in many Tech endeavors. But like many startups he underestimated how much effort would be required for each part of the creation, production, and support of the device. So if anything he should have gotten some VC funding, had them bring in a CEO to run the company (e.g. secure good suppliers in China - with all the issues our Tennis Warehouse people have had w devices broken upon arrival, he obviously chose badly) to free him up to keep improving the tech.
 
@onehandbh i agree w your 1st paragraph. I saw the Bloomberg story - the inventor touted it as s game changer.

To the rest of your comment: no. Look him up. He doesn’t need any Physics students’ advice. He’s been very successful in his career in many Tech endeavors. But like many startups he underestimated how much effort would be required for each part of the creation, production, and support of the device. So if anything he should have gotten some VC funding, had them bring in a CEO to run the company (e.g. secure good suppliers in China - with all the issues our Tennis Warehouse people have had w devices broken upon arrival, he obviously chose badly) to free him up to keep improving the tech.
Oh. I didn’t know about his past endeavors.

Makes it all the more surprising to me that he thought he could squeeze blood out of a stone(his camera and equipment limitations)
 
The device has some design limitations .I opened up a battery and it seems pretty fragile.

Does anyone have advice on opening up the case? Both my cameras popped out when it fell on my carpet. I am out until this gets fixed.
 
The device has some design limitations .I opened up a battery and it seems pretty fragile.

Does anyone have advice on opening up the case? Both my cameras popped out when it fell on my carpet. I am out until this gets fixed.
Pop open the device and reconnect the cameras. From the updated site, looks like v1.1 has changes to prevent the cameras from popping out.
 
Anyone updated to: Beta2 2018-04-04?
Single Match Play is active. with on screen coin toss....
how does it know which side Player serves first to start?

some lines calls were off the mark.
 
I upgraded to the 4-4-2018 firmware. Works well. Love seeing the MPH speed of shots. But I cannot login to the app on my iphone to see the stats. I enter the email, password, hit "log in" and nothing happens. I tried closing the app, and totally re-registering - nothing works.

Honestly if the stats are the only thing that comes of this device, it would still be really cool for $200 USD.
 
I cannot log in either.

Also, the match mode has a coin flip but how does it know which side is serving? I tried to play a match with it but gave up after a game because it was calling good serves out .

I could not try the smart watch because that app seems to be broken.
 
I cannot log in either.

Also, the match mode has a coin flip but how does it know which side is serving? I tried to play a match with it but gave up after a game because it was calling good serves out .

I could not try the smart watch because that app seems to be broken.
I did a support ticket. They told me to uninstall the app on my phone and reinstall it and login. I did that and now it is all working.
 
Hi,
so does anyone has a device with new enclosure and with latest firmware and can tell me how is with the accuracy? But not for line calling but for analytics. So, do all the shots (bounces) get notated and how accurate the notations are? Because some months ago on a post comparing accuracy with zenniz it could be seen that the accuracy was not good - the width of the ball bounces was to narrow.

Thanks for the help.
 
Hi,
so does anyone has a device with new enclosure and with latest firmware and can tell me how is with the accuracy? But not for line calling but for analytics. So, do all the shots (bounces) get notated and how accurate the notations are? Because some months ago on a post comparing accuracy with zenniz it could be seen that the accuracy was not good - the width of the ball bounces was to narrow.

Thanks for the help.
I didn't use zennizh so I may be off here. My opinion, Zenniz has more going for it-- multiple sensors and calibration, better vantage point, etc. Also price:: 3000euros.
I expect it to be a lot more accurate than in/out.

In My opinion, in/out for stats and analysis is adequate. For me, I don't need the precision of where my shots are etc. A few cm or couple inches is ok by me. I'm looking at relative improvement. For example, I played an opponent and uses the inout data as my baseline. Every time I played him ( often adjusting my game plan), I get inout stats that I compare with my earlier stats. Given me lots of insights.
That said, the software is still buggy and the battery life sucks.
 
I didn't use zennizh so I may be off here. My opinion, Zenniz has more going for it-- multiple sensors and calibration, better vantage point, etc. Also price:: 3000euros.
I expect it to be a lot more accurate than in/out.

In My opinion, in/out for stats and analysis is adequate. For me, I don't need the precision of where my shots are etc. A few cm or couple inches is ok by me. I'm looking at relative improvement. For example, I played an opponent and uses the inout data as my baseline. Every time I played him ( often adjusting my game plan), I get inout stats that I compare with my earlier stats. Given me lots of insights.
That said, the software is still buggy and the battery life sucks.
Yes but is the precision in ranges of cm or inches or is it in meters? Because in centimeters is fine, but if it is in meters then it is hard to measure improvement. Tnx
 
My experience is that it isn't off by much. It's actually accurate in most spots on a well lit court and only off certain spots. I never felt that it was off beyond a foot or so even in those scenarios. Your mileage may vary.
 

eelhc

Hall of Fame
+/- 1 foot accuracy

Do the people who make bad line calls usually make calls outside of 1 foot range?
I've seen it happen...

Again... this device is only good if both players agree to use it and stick to the calls it makes. Ain't ever going to happen with a guy who hooks calls.

Hawkeye may not be perfect either but the HW/SW/challenge system works because it's built into the rules everyone abides by.

Even if the In/Out device was perfect... it's a long ways from gaining universal acceptance as a line calling device.

In the meantime, I'm enjoying the stats.
 
Do not forget, it also has video replay for disputes. I would imagine accuracy improves if you have 2 devices. Has anyone tried 2 devices? Sun may be an issue for that it the morning and afternoon .
 
The recently firmwares brings the accuracy to inches for me.
How were you able to determine that the accuracy was within inches?
Did you have a high speed camera recording the same landing spot?
I don't have high speed camera but clay courts ;) and it's better for doing accuracy test.
I made such tests with the 2018-02-02 firmware and with the latest 2018-05-05.
And there is great improvements between them.

With the 2018-05-05 firmware, the average difference, in cm, between the position computed from the device and real position is 6.6cm in the length of the court and 3.8cm in the width. Even better if the bounce is in the half court near the device : 4.5cm in the length and 2.2cm in the width (against 15cm and 18.3cm with 2018-02-02 firmware).

The results are detailed in the two following pictures :

  • test with 2018-05-05 firmware
  • test with 2018-02-02 firmware
And my process is detailed in the following picture:


The accuracy is starting to be very interesting. And don't forget that there is another process for linecall : pure computer vision to "see" if the ball is "before" or "after" the line.
 
Thanks for this great analysis amalguy. I did something similar but not as sophisticated and observed like you about 2cm difference in the width, on the sidelines at midcourt
How many devices were you using? It would be very interesting to determine the difference in accuracy between 1 and 2 devices. I assume the second device is only relevant for improving sideline calls
Anyway, these are still significant margins of error. I find the errors are large enough to be seen with the player's eyes on most parts of the court so if you have 2 honest players the system is practically unusable for close line calls during a match (too many bad calls). This is obviously even more of an issue on clay - I'm sure we're all very curious to see how much more improvement we'll get wit the next firmware
 
Can a scientist/engineer shed light on what may be causing the ball contact point accuracy issues described by amalguy above and comment on my amateur logic here:


Assuming perfect court calibration (ie good light conditions etc..) it seems that


1) the device could be struggling to either determine the exact moment the ball makes contact with the court, in other words the frame rate of the camera may not be fast enough to capture the exact frame corresponding to the contact point which would mean that an estimate of the contact point would have to be calculated based on the ball trajectory/speed /position in previous or later frames. Basic calculation to illustrate:

150km /h ball if filmed at 240 frames per second (I think this is GoPro equivalent) means 1 frame covers a distance of 17.3 cm which is much bigger than a ball contact point. However in amalguy’s experiment, he’s just throwing the ball so slower speed would mean much smaller distance per frame rate so doesn’t explain the errors... maybe the camera films at much slower speed?


and/or 2) the pixel quality of the camera is a limitation meaning that even if the contact point were captured by the camera, there are not enough pixels in the images to precisely calculate the distance to the lines


Despite this, what still puzzles me is line call mistakes which are very clearly visible in the replay mode (ie you can clearly see a gap between the ball and the line when you zoom in) - surely that should be an easy software fix
 
I don't have high speed camera but clay courts ;) and it's better for doing accuracy test.
I made such tests with the 2018-02-02 firmware and with the latest 2018-05-05.
And there is great improvements between them.

With the 2018-05-05 firmware, the average difference, in cm, between the position computed from the device and real position is 6.6cm in the length of the court and 3.8cm in the width. Even better if the bounce is in the half court near the device : 4.5cm in the length and 2.2cm in the width (against 15cm and 18.3cm with 2018-02-02 firmware).

The results are detailed in the two following pictures :

  • test with 2018-05-05 firmware
  • test with 2018-02-02 firmware
And my process is detailed in the following picture:


The accuracy is starting to be very interesting. And don't forget that there is another process for linecall : pure computer vision to "see" if the ball is "before" or "after" the line.
Awesome work! I've also noticed that since the switch to the pixel line calling algorithm in 2018-03-03 firmware, the ball accuracy has been substantially better!
 
I don't have high speed camera but clay courts ;) and it's better for doing accuracy test.
I made such tests with the 2018-02-02 firmware and with the latest 2018-05-05.
And there is great improvements between them.

With the 2018-05-05 firmware, the average difference, in cm, between the position computed from the device and real position is 6.6cm in the length of the court and 3.8cm in the width. Even better if the bounce is in the half court near the device : 4.5cm in the length and 2.2cm in the width (against 15cm and 18.3cm with 2018-02-02 firmware).

The results are detailed in the two following pictures :

  • test with 2018-05-05 firmware
  • test with 2018-02-02 firmware
And my process is detailed in the following picture:


The accuracy is starting to be very interesting. And don't forget that there is another process for linecall : pure computer vision to "see" if the ball is "before" or "after" the line.
Very nice work!
 
I don't have high speed camera but clay courts ;) and it's better for doing accuracy test.
I made such tests with the 2018-02-02 firmware and with the latest 2018-05-05.
And there is great improvements between them.

With the 2018-05-05 firmware, the average difference, in cm, between the position computed from the device and real position is 6.6cm in the length of the court and 3.8cm in the width. Even better if the bounce is in the half court near the device : 4.5cm in the length and 2.2cm in the width (against 15cm and 18.3cm with 2018-02-02 firmware).

The results are detailed in the two following pictures :

  • test with 2018-05-05 firmware
  • test with 2018-02-02 firmware
And my process is detailed in the following picture:


The accuracy is starting to be very interesting. And don't forget that there is another process for linecall : pure computer vision to "see" if the ball is "before" or "after" the line.
Very interesting - how many devices are you using? Would be good to know if a second device makes a material difference
 
However in amalguy’s experiment, he’s just throwing the ball so slower speed would mean much smaller distance per frame rate
In fact, I throwed the ball with hand for my old tests with 2018-02-02 firmware. For my recents tests with 2018-05-05 firmware, I use my racket with an average speed of 95 km/h.

It would be very interesting to determine the difference in accuracy between 1 and 2 devices. I assume the second device is only relevant for improving sideline calls
Very interesting - how many devices are you using? Would be good to know if a second device makes a material difference
I have two devices and made some tests with the two paired with 2018-02-02 firmware. The only communication between them is to hold the case of one missing a bounce. In this case, the other transmits his stats.
So each device, in same record period, contains stats about the same scope of bounces. And for bounces seen by the 2 devices, the stats (position, speed and height over the net) are differents for the 2 devices. And stored as is.
No IA to find the device who have the more accurate data. A first simple "IA" would be to use the sats provided by the device nearest the rebound.
And at the end of the session, you get two differents stats session, one for each device.
I'm in relation with Gregoire to get a merged session with best line calling/stats from the two devices. But no plan for this soon.
But there is improvement in the communication between paired devices in the next release 2018-06-06. Then perhaps good news soon.
 
In fact, I throwed the ball with hand for my old tests with 2018-02-02 firmware. For my recents tests with 2018-05-05 firmware, I use my racket with an average speed of 95 km/h.



I have two devices and made some tests with the two paired with 2018-02-02 firmware. The only communication between them is to hold the case of one missing a bounce. In this case, the other transmits his stats.
So each device, in same record period, contains stats about the same scope of bounces. And for bounces seen by the 2 devices, the stats (position, speed and height over the net) are differents for the 2 devices. And stored as is.
No IA to find the device who have the more accurate data. A first simple "IA" would be to use the sats provided by the device nearest the rebound.
And at the end of the session, you get two differents stats session, one for each device.
I'm in relation with Gregoire to get a merged session with best line calling/stats from the two devices. But no plan for this soon.
But there is improvement in the communication between paired devices in the next release 2018-06-06. Then perhaps good news soon.
Thanks, very useful info! Do you think that camera speed/frame rate per second is the main constraint in this device?...meaning that ball impact point often has to be estimated instead of 'really' filmed?..I read in the InOut support section that recording video in slow motion is done at 120 fps..so I assume that is the fastest speed of recording...and therefore the basic calculation I showed in my post on May 23 is in fact too optimistic?
 
Thanks, very useful info! Do you think that camera speed/frame rate per second is the main constraint in this device?...meaning that ball impact point often has to be estimated instead of 'really' filmed?..I read in the InOut support section that recording video in slow motion is done at 120 fps..so I assume that is the fastest speed of recording...and therefore the basic calculation I showed in my post on May 23 is in fact too optimistic?
I did the math too, and came up with a similar result. What is interesting is that the human eye/brain can't see or distinguish faster than about 120 frames per second (or somewhere around there depends on training, and periphery, etc.). So the camera is at least equal to our eyes. But, we have intelligence to make 'calculations' about where the ball bounces. The In/Out algorithm is suppose to use AI. That's where the magic lies, if In/Out ever gets the AI correct.
 
150km /h ball if filmed at 240 frames per second (I think this is GoPro equivalent) means 1 frame covers a distance of 17.3 cm which is much bigger than a ball contact point.
Thanks, very useful info! Do you think that camera speed/frame rate per second is the main constraint in this device?...meaning that ball impact point often has to be estimated instead of 'really' filmed?..I read in the InOut support section that recording video in slow motion is done at 120 fps..so I assume that is the fastest speed of recording...and therefore the basic calculation I showed in my post on May 23 is in fact too optimistic?
I think that the framerate of the video used to calculate the position of ball's bounces is 120fps.
If we assume that the framerate used for the bounce's replay video on the device is the same as the one used for calculation, which seems for me a good assumption, the following picture demonstrates that the framerate is 120fps

In this picture, we can see :
  • at top the frames of the bounce's replay video recorded on my phone mirored with the device. The recording was done with "AZ Screen Recorder"
  • at bottom, the video recorded by the device with record options set to 120 fps slow motion
The slow motion video is in fact at 95fps not at the expected 120fps. And we can see that the bounce's replay video use the same 6 frames as the 95fps slow motion video + 2 frames not present in slow motion. The 8 frames correspond to a final framerate of 120fps.

With these settings, with no compute of ball trajectory, the maximum reachable accuracy for the calculation of position of my shots at 95km/h would be ~21cm.
But the real mesured accuracy on clay court are 4.5cm in the length and 2.2cm in the width.
So In/Out seems to do a very good job in computing the ball trajectory path to find the position of the ball at timestamps not covered by video frames.
In this reflection, I suppose that the real accuracy would be the same on hard courts than on clay courts.
 
I think that the framerate of the video used to calculate the position of ball's bounces is 120fps.
If we assume that the framerate used for the bounce's replay video on the device is the same as the one used for calculation, which seems for me a good assumption, the following picture demonstrates that the framerate is 120fps
.
the devices records at 720p with 30fps ; thats the only way how the device can replay so many shots, the replay cam Shows only black and White colour to save space
what else should it have?

120fps? do you know how much storage capacitiy is needed for 120fps Videos in a high resolution?
 
Top