> I wanted to cook venison from scratch, which meant learning to shoot, which meant keeping track of my progress, which meant porting a 2012 OpenCV paper and training a state-of-the-art computer vision model, which meant the dinner took a bit longer than expected.
Yup. Or deer shaving, in this case. The punchline is he never actually got round to shooting a deer.
I do think he understated the difficulty of the hunt itself. He's planning to use the "supervision" rule to avoid needing his own firearm license, and male deer are indeed unlicensed for shooting (but not female deer!). Then you have to find one. He's right that they have reached "pest" status, since humans killed off the wolves. Every now and again someone suggests reintroducing the wolves, to cull the deer (and occasional tourists).
The open terrain (because the deer eat saplings) can make it easier. I have a great photo somewhere of a single majestic deer which I just happened to see from the road when I had my telephoto lens with me and mounted on the camera. I've even once seen a deer in Edinburgh itself, along a railway cutting.
My USPSA rank is public: I'm terrible with pistols. I haven't shot in competition for over a decade. This is the kind of project that tickles a couple of my nerves and might get me back to the range.
Scoring is based on the outermost ring, rather than the innermost ring?
Huh. I'd have expected it to be based on the center, but I guess the goal is "it must be entirely within this ring to count" rather than just "I hit this ring".
I think it depends on the discipline, NSRA .22 in the post uses the outermost edge, but ISSF (Olympic rifle/pistol, for example) uses the innermost edge.
For comment reading edification, there are already electronic scoring targets for shooting.[0]
They use wave detection from each corner - either air/sound or via the target backing - to triangulate and with modern electronics can be quite accurate.
It's nice from an audience point of view to be able to see the results of each shot almost immediately. Kinda like watching snooker championships.
This approach is novel however and has other pros and cons.
There are also targets with fluorescent backgrounds and special black paint that flakes off near the holes. There is a limit to how many holes you can see in the target but it is way better than plain paper.
That has to have been an intentional joke; Greggs' aren't good, just cheap and ubiquitous, and almost any proper hand-made-on-site bakery will beat them.
I've been building a similar piece of software but with vibe coding. It's to the point that I'm using gauge blocks to measure the precise scoring ring dimensions and then using various warping techniques to get the photo to map precisely. In a weekend I've been able to get it to sub pixel accuracy.
Um... No. An american 22 can be very slightly smaller. American-invented calibers are measured to the depth of the grooves in a rifled barrel. The rest of the world measures to the flat parts between the grooves. So no, it is not obvious how wide a bullet is.
And beware the plural. If someone (usually a salty navy person) says that a gun is "50 calibers" he means something completely different than a "50 caliber".
Legacy and marketing have as much to do with it as local variations in how bores are actually measured do.
All the .38s and 9mms of the world are just slight variations on .36" round ball, .44 caliber pistols are generally .429", there's a .45" pistol caliber labelled .460 (.454 also counts), .50 BMG is actually .510", calibers claimed to be "7.62mm" use either a .308" or .311" projectile depending on the country of origin and sometimes not even then (France and Switzerland call this size 7.5mm, Argentina called this 7.65mm, Japan called it 7.7mm, the British called it .303), "8mm" can be either a .318" or .323" projectile, .32s are all .312" diameter, but one cartridge that uses this same projectile labels it as .30 and another .327.
The same 5.7mm projectile (.224") is used in cartridges that claim to have a diameter of .220, .221, .222, .223, .224, .225, 5.6mm, 5.56mm, and 5.7mm.
.277" projectiles are used in cartridges that call themselves 6.8mm, .270, .277; same thing with .284" projectiles used in cartridges that call themselves 7mm and .280.
Not just diameter, the bullet itself is identical. The cartridge is longer with more powder in a .357, which makes it a good bit more powerful in practice (2 to 3 times as much energy).
The explanation for the caliber discrepancy is halfway interesting:
"Despite its name, the caliber of the .38 Special cartridge is actually .357 inches (36 caliber/9.07 mm), with the ".38" referring to the approximate diameter of the loaded brass case. This came about because the original 38-caliber cartridge, the .38 Short Colt, was designed for use in converted .36-caliber cap-and-ball Navy revolvers, which had untapered cylindrical firing chambers of approximately 0.374-inch (9.5 mm) diameter that required heeled bullets, the exposed portion of which was the same diameter as the cartridge case."
It is more complex. A softer bullet can get wider durong firing. And the barrel can expand slight also. This is why caliber invention isnt as simple as just picking a size.
Of all the things one can automate in this whole journey - he chose the ring counting on the shooting range? I don't get it.
I totally see the programming challenge there, but it's in no substantial way making the journey any easier. Any somewhat working human brain can count this quite quickly and then move on with other things.
Because he kept hitting his head on the low ceiling beams as he walked over to look at his target.
If he had been shooting at an outdoor range, or even an indoor range with a higher ceiling, he probably wouldn't have been pushed to automate the process.
Counting rings is easy indeed, but scoring borderline shots without a scoring gauge is not, because the visible bullet hole is often smaller than the bullet itself.
But why would he care about this millimeter precision? His objective is not to participate in the Olympics but to shoot deer. He wants to improve general shooting abilities, not sub-millimeter accuracy. If he now and then counts a ring wrong, then what's the problem? That's what I don't get.
Now that the software exists, one can use it from a mounted camera and provide immediate scoring. No need to wait for the human and the target to be in proximity.
> I wanted to cook venison from scratch, which meant learning to shoot, which meant keeping track of my progress, which meant porting a 2012 OpenCV paper and training a state-of-the-art computer vision model, which meant the dinner took a bit longer than expected.
Procrastination level: Ultimate
yak shaving
Yup. Or deer shaving, in this case. The punchline is he never actually got round to shooting a deer.
I do think he understated the difficulty of the hunt itself. He's planning to use the "supervision" rule to avoid needing his own firearm license, and male deer are indeed unlicensed for shooting (but not female deer!). Then you have to find one. He's right that they have reached "pest" status, since humans killed off the wolves. Every now and again someone suggests reintroducing the wolves, to cull the deer (and occasional tourists).
The open terrain (because the deer eat saplings) can make it easier. I have a great photo somewhere of a single majestic deer which I just happened to see from the road when I had my telephoto lens with me and mounted on the camera. I've even once seen a deer in Edinburgh itself, along a railway cutting.
The title is so bad, can't understand what's it about. Atleast around other topics here.
No, it’s a great title. It was curious enough that I clicked and then I read the whole thing.
It’s unlikely I would have done that if it had been something like “Using computer vision to score rifle shooting cards”.
In other words it's a great clickbait title, yet still a bad title.
My USPSA rank is public: I'm terrible with pistols. I haven't shot in competition for over a decade. This is the kind of project that tickles a couple of my nerves and might get me back to the range.
A very nice read and surprising ending. Love the perseverance of the research and the cleanliness of the app.
Scoring is based on the outermost ring, rather than the innermost ring?
Huh. I'd have expected it to be based on the center, but I guess the goal is "it must be entirely within this ring to count" rather than just "I hit this ring".
I think it depends on the discipline, NSRA .22 in the post uses the outermost edge, but ISSF (Olympic rifle/pistol, for example) uses the innermost edge.
This ... Is beautiful
For comment reading edification, there are already electronic scoring targets for shooting.[0]
They use wave detection from each corner - either air/sound or via the target backing - to triangulate and with modern electronics can be quite accurate.
It's nice from an audience point of view to be able to see the results of each shot almost immediately. Kinda like watching snooker championships.
This approach is novel however and has other pros and cons.
[0] https://en.wikipedia.org/wiki/Electronic_scoring_system
There are also targets with fluorescent backgrounds and special black paint that flakes off near the holes. There is a limit to how many holes you can see in the target but it is way better than plain paper.
> Best gourmet pastries this side of the pond (and they also do doughnuts!).
Haha no, not even close. Saddler's Forfar Bridies take that prize.
That has to have been an intentional joke; Greggs' aren't good, just cheap and ubiquitous, and almost any proper hand-made-on-site bakery will beat them.
Wasn't sure what to expect when I clicked this link.
I've been building a similar piece of software but with vibe coding. It's to the point that I'm using gauge blocks to measure the precise scoring ring dimensions and then using various warping techniques to get the photo to map precisely. In a weekend I've been able to get it to sub pixel accuracy.
>> .22 bullet is 0.22" across (duh)
Um... No. An american 22 can be very slightly smaller. American-invented calibers are measured to the depth of the grooves in a rifled barrel. The rest of the world measures to the flat parts between the grooves. So no, it is not obvious how wide a bullet is.
And beware the plural. If someone (usually a salty navy person) says that a gun is "50 calibers" he means something completely different than a "50 caliber".
https://en.wikipedia.org/wiki/Caliber
Wow, I had no idea. The ones we had at the range were 0.22LR and the boxes are marked 5.7 mm which is also not precisely 0.22".
Yeah, given the nominal precision it's surprising how far off some of the numbers are. A .38 is the same diameter as a .357.
Legacy and marketing have as much to do with it as local variations in how bores are actually measured do.
All the .38s and 9mms of the world are just slight variations on .36" round ball, .44 caliber pistols are generally .429", there's a .45" pistol caliber labelled .460 (.454 also counts), .50 BMG is actually .510", calibers claimed to be "7.62mm" use either a .308" or .311" projectile depending on the country of origin and sometimes not even then (France and Switzerland call this size 7.5mm, Argentina called this 7.65mm, Japan called it 7.7mm, the British called it .303), "8mm" can be either a .318" or .323" projectile, .32s are all .312" diameter, but one cartridge that uses this same projectile labels it as .30 and another .327.
The same 5.7mm projectile (.224") is used in cartridges that claim to have a diameter of .220, .221, .222, .223, .224, .225, 5.6mm, 5.56mm, and 5.7mm.
.277" projectiles are used in cartridges that call themselves 6.8mm, .270, .277; same thing with .284" projectiles used in cartridges that call themselves 7mm and .280.
> A .38 is the same diameter as a .357.
Not just diameter, the bullet itself is identical. The cartridge is longer with more powder in a .357, which makes it a good bit more powerful in practice (2 to 3 times as much energy).
The explanation for the caliber discrepancy is halfway interesting:
"Despite its name, the caliber of the .38 Special cartridge is actually .357 inches (36 caliber/9.07 mm), with the ".38" referring to the approximate diameter of the loaded brass case. This came about because the original 38-caliber cartridge, the .38 Short Colt, was designed for use in converted .36-caliber cap-and-ball Navy revolvers, which had untapered cylindrical firing chambers of approximately 0.374-inch (9.5 mm) diameter that required heeled bullets, the exposed portion of which was the same diameter as the cartridge case."
https://en.wikipedia.org/wiki/.38_Special
Won't a fired .22 bullet be sized to the depth of the grooves due to obturation?
It is more complex. A softer bullet can get wider durong firing. And the barrel can expand slight also. This is why caliber invention isnt as simple as just picking a size.
Of all the things one can automate in this whole journey - he chose the ring counting on the shooting range? I don't get it.
I totally see the programming challenge there, but it's in no substantial way making the journey any easier. Any somewhat working human brain can count this quite quickly and then move on with other things.
Really, I don't get it.
Because he kept hitting his head on the low ceiling beams as he walked over to look at his target.
If he had been shooting at an outdoor range, or even an indoor range with a higher ceiling, he probably wouldn't have been pushed to automate the process.
Counting rings is easy indeed, but scoring borderline shots without a scoring gauge is not, because the visible bullet hole is often smaller than the bullet itself.
But why would he care about this millimeter precision? His objective is not to participate in the Olympics but to shoot deer. He wants to improve general shooting abilities, not sub-millimeter accuracy. If he now and then counts a ring wrong, then what's the problem? That's what I don't get.
> His objective is not to participate in the Olympics but to shoot deer.
Where do you see that?
The article is about someone in Scotland who took up marksmanship as a hobby.
Now that the software exists, one can use it from a mounted camera and provide immediate scoring. No need to wait for the human and the target to be in proximity.