This is a cool idea, nice work. I spent some time as an assistant college coach in the NCAA and the amount of time spent reviewing film and capturing key plays/schemes etc is huge. I do some computer vision and always wondered if it would ever be accurate enough to skip over dribbling up the courts, time outs etc so scouts could just get to the important parts of the film—it looks like you are on track with what you claim it can do. I might sign up for the beta... good luck!
Yes please try it out, I'd love to know what you think! My email is mike@hooper.gg if you want to chat more about basketball and tech. I would be curious to learn what you have tried so far and what was important to you as a coach
This is awesome, thanks for sharing! Hooper sounds like a fantastic app for pickup basketball players. I love the idea of tracking actual gameplay and creating highlights. The ability to sync recordings and differentiate shots is impressive. Can't wait to try it out and see how it tracks my games. Great work!
Amazing. Pitched something like this to a friend a few days ago and its awesome that you've built it. Headed to a game now and down to use it and give feedback (im in two groups and both have people setting up phones to record). Now add in an AI ref so people can stop arguing for 10 minutes over wether someone stepped out of bounds or not in a pickup game.
Used it this morning and had a few quick thoughts:
1) Love the feed, realized its nice to see other random games i have no connection to. Wouldn't have thought it would be entertaining but its great.
2) Export video to camera roll: im sure some folks will want to export the whole thing (update: now i see that you can do it when you full screen it, may want to move that to the menu on the video processing screen).
3) Creating sharable clips that are branded can definitely help grow the app. Seems like you may be able to identify types of highlights, so mashing them up, with some graphic overlays and voiceovers (think play by play with elevenlabs). Have a few folks who work in this space for sports teams that may be able to give some guidance.
4) Turning the screen off while recording is a brilliant feature and is reassuring
5) Realtime scoring is likely harder but would be a killer feature.
This is awesome - thanks for trying it out and for the feedback! I really appreciate it and it goes a long way to help us make the app better.
+1 on the graphic overlays and voiceovers. I personally want to add Mike Breen's Bang bang voice over as an option haha. Also check out mixtapes on your profile page if you haven't yet. You can mashup highlights that way from different recorded sessions.
+1 on realtime scoring and processing. Right now, a recording takes ~30min to process. We are trying to reduce this over time as much as we can until it is close to live.
This is very interesting and well executed for the initial release. Good luck!
One thing I found odd was that the default/main screen in the app is a feed. That feed is just short videos without any of the game data the app captures. Perhaps an overlay would showoff the app as well as the skills. The videos by themselves don't add much differentiated value.
Yeah, that could be cool -- we played around with some designs on showing the leading scorer or some ranking in each feed card as well. I can't say we've cracked it yet but I like the idea of an overlay.
I love this! looked into doing a similar project, you're competing against Hudl but using the phone instead of custom hardware (always preferred). Highlight segmentation may be a challenge with SOTA cv methods, but there are lot of directions you can go in.
Yeah Hudl is definitely a beast in the space though I think of them as almost more focused on orgs and teams rather than consumers.
Re: highlight segmentation was and still is a challenge that we work on. In the beginning we had a hard time dealing with false positives when our models thought a shot was in but it wasn't. This has gotten better over time with more data and is hovering near 92% acc these days but obv not 100%. But i'm optimistic as data grows and the sota methods get better, this isn't the hardest ml problem in the world
Yeah! We are planning to add blocks, steals, and rebounds in the coming months. I think fouls is still a reach goal right now (hard even for humans) but maybe one day.
Re: capturing the whole court, I think we are considering supporting panning of the camera (in addition to syncing). We focus pretty heavily on making sure things work with just a normal phone so it's more accessible to everyone. We are definitely starting bottom up but maybe one day we will get to more high tech hardware.
In relation to user experience on first try...I just wanted to test it out for 30 second to see if it's worth keeping. I haven't tested now, will have to wait until I get an hour free later because there are a few roadblocks.
- You require email and password registration and to click a verification link. Not too bad
- When first launched the app asks you to allow mic permissions by clicking settings. However in Hooper settings on Android there is no mic permissions option. Bad.
- You require 5gb free to use the product at all. Understandable given storage is required for video but how about reducing that so users can do like a 30-second test for first use? That way they don't have to spend an hour going through their phone's media to see what they want to keep and what they can delete.
You'd be amazed at how many people (like me lol) have almost full storage on their device most of the time.
Edit: oh....it's not a setting for mic permissions, it's a general permissions setting and if you touch that you can add additional permissions. Based on the text of the popup I hadn't guessed that was the right place to look and I hadn't realised it was possible to add permissions by touching that area of the settings menu.
Are you using some sort of naive profanity filtering for profile names? I couldn't create an user name with "assist" in it due to "ass" being included?
Yeah, it would be super cool if all sports had things like this, i think it really enriches the player's experience. I know in tennis there is swingvision so definitely seems like things are headed that way.
We ended up having to collect our own data and train our own deep learning models for detecting balls, players, court, actions, and tracking. But this was because we needed to make things work almost all the time to achieve our desired accuracy. You should try out off-the-shelf things like YOLO, Segment Anything, and Detectron and then after that you probably need to invest more on data collection and model training.
Adidas made some soccer balls with built-in sensors that can provide data in almost real-time to your phone (ie. take a shot and review ball speed, rotation, point of strike etc...). Video: https://www.youtube.com/watch?v=mpIOnnU1R_o&t=119s (unfortunately no english subtitles).
Also, the technology was used in their latest Euro 2024 ball - the Fussballliebe - to help video referees and their decisions.
Very awesome, congrats on your release! I know of a similar golf app that made absolute bank, so if there's a niche and you can tap into it, you guys will likely do quite well.
Here are some i looked at - SwingU, Ace Trace, Swing Profile, Sportsbox AI. Overall seems like two kinds of golf apps: (1) more micro- to track and grade your swing, and (2) more macro- something to track your progress over a course
92% accuracy is really good! Is this just action recognition and player identity at the moment or are you actually generating full tracking data under the hood?
This is a cool idea, nice work. I spent some time as an assistant college coach in the NCAA and the amount of time spent reviewing film and capturing key plays/schemes etc is huge. I do some computer vision and always wondered if it would ever be accurate enough to skip over dribbling up the courts, time outs etc so scouts could just get to the important parts of the film—it looks like you are on track with what you claim it can do. I might sign up for the beta... good luck!
Yes please try it out, I'd love to know what you think! My email is mike@hooper.gg if you want to chat more about basketball and tech. I would be curious to learn what you have tried so far and what was important to you as a coach
This is awesome, thanks for sharing! Hooper sounds like a fantastic app for pickup basketball players. I love the idea of tracking actual gameplay and creating highlights. The ability to sync recordings and differentiate shots is impressive. Can't wait to try it out and see how it tracks my games. Great work!
Yeah! Excited to see what you think of it. Making our tracking algos and shot classifiers better is something we think about a lot.
Amazing. Pitched something like this to a friend a few days ago and its awesome that you've built it. Headed to a game now and down to use it and give feedback (im in two groups and both have people setting up phones to record). Now add in an AI ref so people can stop arguing for 10 minutes over wether someone stepped out of bounds or not in a pickup game.
Used it this morning and had a few quick thoughts:
1) Love the feed, realized its nice to see other random games i have no connection to. Wouldn't have thought it would be entertaining but its great.
2) Export video to camera roll: im sure some folks will want to export the whole thing (update: now i see that you can do it when you full screen it, may want to move that to the menu on the video processing screen).
3) Creating sharable clips that are branded can definitely help grow the app. Seems like you may be able to identify types of highlights, so mashing them up, with some graphic overlays and voiceovers (think play by play with elevenlabs). Have a few folks who work in this space for sports teams that may be able to give some guidance.
4) Turning the screen off while recording is a brilliant feature and is reassuring
5) Realtime scoring is likely harder but would be a killer feature.
Love this, all the best on the development!
This is awesome - thanks for trying it out and for the feedback! I really appreciate it and it goes a long way to help us make the app better.
+1 on the graphic overlays and voiceovers. I personally want to add Mike Breen's Bang bang voice over as an option haha. Also check out mixtapes on your profile page if you haven't yet. You can mashup highlights that way from different recorded sessions.
+1 on realtime scoring and processing. Right now, a recording takes ~30min to process. We are trying to reduce this over time as much as we can until it is close to live.
https://youtu.be/SUvfKX7yXLo?si=0C1XTGOa2NTUEaHy
oh dang! Thanks for the feature- this was sick to watch. Appreciate you putting the time in to give your thoughts.
This is very interesting and well executed for the initial release. Good luck!
One thing I found odd was that the default/main screen in the app is a feed. That feed is just short videos without any of the game data the app captures. Perhaps an overlay would showoff the app as well as the skills. The videos by themselves don't add much differentiated value.
Yeah, that could be cool -- we played around with some designs on showing the leading scorer or some ranking in each feed card as well. I can't say we've cracked it yet but I like the idea of an overlay.
This combines three of my greatest passions - basketball, computer vision, and analytics. I love it! Thanks for sharing :)
Thanks for the comment! Would love to know what you think and how we can make the app better :)
I love this! looked into doing a similar project, you're competing against Hudl but using the phone instead of custom hardware (always preferred). Highlight segmentation may be a challenge with SOTA cv methods, but there are lot of directions you can go in.
Yeah Hudl is definitely a beast in the space though I think of them as almost more focused on orgs and teams rather than consumers.
Re: highlight segmentation was and still is a challenge that we work on. In the beginning we had a hard time dealing with false positives when our models thought a shot was in but it wasn't. This has gotten better over time with more data and is hovering near 92% acc these days but obv not 100%. But i'm optimistic as data grows and the sota methods get better, this isn't the hardest ml problem in the world
Cool but offence is only half the story! Any plans to add defensive stats or stuff like turnovers / steals etc?
Also any plans to make it work from camera footage e.g. A 360 camera that can capture the whole court at once instead of syncing?
Yeah! We are planning to add blocks, steals, and rebounds in the coming months. I think fouls is still a reach goal right now (hard even for humans) but maybe one day.
Re: capturing the whole court, I think we are considering supporting panning of the camera (in addition to syncing). We focus pretty heavily on making sure things work with just a normal phone so it's more accessible to everyone. We are definitely starting bottom up but maybe one day we will get to more high tech hardware.
Re:fouls.
At my pickups courts we call our own fouls by saying Foul. Perhaps the mic can pickup key words matched the play stopping.
Oh neat, yeah we could some speech to text thing or maybe even some like visual signal thing to call fouls.
Very cool of you to take on board feedback.
In relation to user experience on first try...I just wanted to test it out for 30 second to see if it's worth keeping. I haven't tested now, will have to wait until I get an hour free later because there are a few roadblocks. - You require email and password registration and to click a verification link. Not too bad - When first launched the app asks you to allow mic permissions by clicking settings. However in Hooper settings on Android there is no mic permissions option. Bad. - You require 5gb free to use the product at all. Understandable given storage is required for video but how about reducing that so users can do like a 30-second test for first use? That way they don't have to spend an hour going through their phone's media to see what they want to keep and what they can delete.
You'd be amazed at how many people (like me lol) have almost full storage on their device most of the time.
Edit: oh....it's not a setting for mic permissions, it's a general permissions setting and if you touch that you can add additional permissions. Based on the text of the popup I hadn't guessed that was the right place to look and I hadn't realised it was possible to add permissions by touching that area of the settings menu.
Are you using some sort of naive profanity filtering for profile names? I couldn't create an user name with "assist" in it due to "ass" being included?
Awesome! I someday hope to find the time to make something like this for soccer - individual skill drills as well as gameplay analysis.
Care to share any technical details about how your analysis pipeline works? :)
Yeah, it would be super cool if all sports had things like this, i think it really enriches the player's experience. I know in tennis there is swingvision so definitely seems like things are headed that way.
We ended up having to collect our own data and train our own deep learning models for detecting balls, players, court, actions, and tracking. But this was because we needed to make things work almost all the time to achieve our desired accuracy. You should try out off-the-shelf things like YOLO, Segment Anything, and Detectron and then after that you probably need to invest more on data collection and model training.
Adidas made some soccer balls with built-in sensors that can provide data in almost real-time to your phone (ie. take a shot and review ball speed, rotation, point of strike etc...). Video: https://www.youtube.com/watch?v=mpIOnnU1R_o&t=119s (unfortunately no english subtitles).
Also, the technology was used in their latest Euro 2024 ball - the Fussballliebe - to help video referees and their decisions.
Very awesome, congrats on your release! I know of a similar golf app that made absolute bank, so if there's a niche and you can tap into it, you guys will likely do quite well.
Appreciate the kind words!
What's the golf app?
Here are some i looked at - SwingU, Ace Trace, Swing Profile, Sportsbox AI. Overall seems like two kinds of golf apps: (1) more micro- to track and grade your swing, and (2) more macro- something to track your progress over a course
92% accuracy is really good! Is this just action recognition and player identity at the moment or are you actually generating full tracking data under the hood?
This looks pretty sick and the site and promo video is very well done also. Nice job.
Thanks! Appreciate the comment!
This looks great. curious, Is is something similar for tennis and/or ice skating?
Check this product out - https://swing.tennis. I've used it before and it's quite good
Well done, ai_bullish++
Please do surfing next and add it to Surfline
What are you using for video analysis?
Baked our own image models that process each frame to detect objects. Then we added some lightweight video models to detect actions and tracking
sweet
[dead]
[dead]