You guys agree with this guy?
- Thread starter tibimakai
- Start date
user8963
Known around here
problem of hookdown and all youtuber..
they made videos for their fanbase/followers ... looking into the comments you see the problem.
they/he will always make the same videos to prove that they are right because they always show the same arguments.
people who watch this videos just prove that they were right in decision to purchase garbage products, so everyone is happy
hookdown gets prove that he is right because everyone tell him that they like reolink which he recommends to them.
also he prove himself that we are wrong and dahua/hikvision is just overpriced garbage.
this shit is called
"preach to the choir" / "preach to the converted"
people who are not in just see all the comments and think that reolink is good.
thatswhy youtube recomendations always suck. you have to crawl through the comments to find real facts (if he didnt delete it).
after months or maybe years you see absolute NO progress in his videos.
i have no idea how many time he invest in the videos, but if you compare so many cameras there can be many faults. maybe you forget to turn off WDR and just have awful image. or you have so much footage that you cant remember which one was which ,... he never talk about any settings. no one knows if he ever touch any settings on reolink cameras.. its just stupid.
they made videos for their fanbase/followers ... looking into the comments you see the problem.
they/he will always make the same videos to prove that they are right because they always show the same arguments.
people who watch this videos just prove that they were right in decision to purchase garbage products, so everyone is happy
hookdown gets prove that he is right because everyone tell him that they like reolink which he recommends to them.
also he prove himself that we are wrong and dahua/hikvision is just overpriced garbage.
this shit is called
"preach to the choir" / "preach to the converted"
people who are not in just see all the comments and think that reolink is good.
thatswhy youtube recomendations always suck. you have to crawl through the comments to find real facts (if he didnt delete it).
after months or maybe years you see absolute NO progress in his videos.
i have no idea how many time he invest in the videos, but if you compare so many cameras there can be many faults. maybe you forget to turn off WDR and just have awful image. or you have so much footage that you cant remember which one was which ,... he never talk about any settings. no one knows if he ever touch any settings on reolink cameras.. its just stupid.
dudemaar
Known around here
For a fair and honest review, we should compare and review his $85 reodink camera to ipcamtalk's $84 IPC-T2431T-AS camera.
ipcamtalk.com
REVIEW: Budget 4MP Turret Comparison - IPC-T2431T-AS vs Chinese HDW4433C-A
Many folks on a budget tend to grab Chinese region cameras off of Amazon because, well they're less expensive, and some simply arent up to speed on the lack of firmware support and other issues. One of the most popular of these Chinese region models is the 4433C-A, a 4MP turret. The 4433C-A...
Enough said. My #1 security concern is at night where prowlers or criminal people are more likely to do their evils.
KenAllen15
Getting the hang of it
Hey @mat200 can you explain like I'm five?
I then took that into photoshop and cropped out his head/face
that resulting face picture is 39px x 52px which is 2028 pixels, which seems like a lot more than 100 ppf. Where am I messing up here?
****edit: maybe it's pixels across the face? or from chin to forehead?
I went into his raw footage video and downloaded the frame from the RLC811A at 40ft.
I then took that into photoshop and cropped out his head/face
that resulting face picture is 39px x 52px which is 2028 pixels, which seems like a lot more than 100 ppf. Where am I messing up here?
****edit: maybe it's pixels across the face? or from chin to forehead?
KenAllen15
Getting the hang of it
Maybe that dude just has a skinny head, but I don't get to 100px across the face until he is only 10ft away:
@KenAllen15 - you are looking at the # pixels, not the per foot part that is the pf in ppf....
PPF is calculated as (Horizontal Pixel Count of sensor)/(field of view width distance).
Since it is Reolink, we will let them explain it LOL:
PPF is calculated as (Horizontal Pixel Count of sensor)/(field of view width distance).
Since it is Reolink, we will let them explain it LOL:
Can 1080p IP Security Cameras Identify License Plate and People - Reolink Blog
Can 1080p IP security cameras identify cars/vehicles license plates and the facial features of people? Why not just jump into this article to get the most perfect answer?
reolink.com
KenAllen15
Getting the hang of it
doesn't the "f" in ppf stand for "face" as in pixels per face?@KenAllen15 - you are looking at the # pixels, not the per foot part that is the pf in ppf....
PPF is calculated as (Horizontal Pixel Count of sensor)/(field of view width distance).
Since it is Reolink, we will let them explain it LOL:
Can 1080p IP Security Cameras Identify License Plate and People - Reolink Blog
Can 1080p IP security cameras identify cars/vehicles license plates and the facial features of people? Why not just jump into this article to get the most perfect answer?reolink.com
mat200
IPCT Contributor
- Joined
- Jan 17, 2017
- Messages
- 13,931
- Reaction score
- 23,234
HI @KenAllen15
Please do let us know if the Cliff Notes needs some edits to cover this better.
As @wittaj mentions "PPF is calculated as (Horizontal Pixel Count of sensor)/(field of view width distance). " ( Pixel Per Foot )
note: in the cliff notes we have a more exact math equation to calculate this... uses arc .. thus degrees and pi are a part of the equation.
Q: "I then took that into photoshop and cropped out his head/face that resulting face picture is 39px x 52px which is 2028 pixels, which seems like a lot more than 100 ppf. Where am I messing up here? "
Any 2D measure would have units like Square Foot. Example say we have a pixel density of 100 ppf on the horizontal and a 100 ppf on the vertical - then 100 ppf * 100 ppf -> gives you a 100,000 pixel per square foot
So you're actually looking at a square unit when you combine the horizontal and vertical ...
Q: Maybe that dude just has a skinny head, but I don't get to 100px across the face until he is only 10ft away
Remember, we have defined the density of the pixels per square foot when you can get enough pixels to be able to statistically get an IDable image. ( identifiable ) Most people's faces are not 1 foot wide .. so naturally there will be less pixels across the horizontal image of the face. ( I don't have the number of pixels across the face memorized, so you'd have to check the references in the cliff notes on what the number is, naturally it will be less than 100 pixels )
Last edited:
KenAllen15
Getting the hang of it
mat200
IPCT Contributor
- Joined
- Jan 17, 2017
- Messages
- 13,931
- Reaction score
- 23,234
Thanks @KenAllen15
Yes, px/face = pixels per face is used.
Also ppf is used .. Pixels Per Foot ( or in the non-USA world .. pixels per meter )
Here's what the research did:
1) Calculate how many pixels ( effective ) you need across someone's face to get a good ID / identifiable image.
This is your pixels per face measure ( notice, horizontal pixels per face is the measure they use .. we assume there's a reasonable horizontal x vertical dimensions taken by the camera ... perhaps even assuming the pixels are basically little "squares".
2) Once you know that, you can calculate the ppf ( pixel per foot ) you need.
take px/face, average width of a face -> calculate ppf
( pixel per face ) / ( average width of a face ) = pixels per measure ( can be inches, cm, foot, meters .. all depends on the unit you are using )
We like to use 100 ppf here ..
3) To determine the ID distance ( that is to the 100 ppf arc ) use the equation in the cliff notes:
Radius = (( Horizontal Res / 100 ppf ) * ( 360 / Angle ))/2*Pi
This will give you the "I" from the DORI calculations.
R is iirc 2x that ..
4) This now gives you something to start your work and planning with ..
Remember, compression algorithms ( which really hit cloud cameras a lot, when they attempt to reduce bandwidth used ), camera tuning parameters, light glare, wdr, hdr, color depth, bright spots, environmental issues ( fog, rain, bugs, spiders, dirt on lens, snow, .. ) all start to reduce the effective pixels.
so what does this all mean?
The reviewer is clearly in the R range for the Reolink 8MP 4mm camera during good lighting conditions.
The Reolink RLC-810a || Resolution: 3840x2160 (8.0 Megapixels) || FOV: Horizontal: 87° Vertical: 44° || ID distance ( Radius ) = ((3840/100)*(360/87)) / 2*Pi = 25.3 feet
If you know the suspect / subject you can in theory recognize them. If you do not, then you should not be able to get a solid positive ID based on the research given a good straight on facial image capture.
SO when the reviewer states at 40 feet neither the Reolink 8MP nor the Dahua 5442 can give a positive ID .. that statement corresponds to the research and math / physics. However, if you really understand DORI you would have known this before even testing the cameras at 40 feet ( day or night ). Thus this leads to the wrong conclusion and is deceptive to anyone new to the topic.
The ID tests should be done within the DORI ranges of the respective cameras, and under different lighting conditions, and the tuning parameters should be shared - as they can have a large image.
This is also a major problem with Reolink - numerous reports that they fail to hold to any tuning parameters attempted by users / owners. ( again, this is a part of the Reolink company being unethical imho by tuning the cameras to get better still image capture to convince unsuspecting buyers they are getting a better camera than the Reolink really is .. )
mat200
IPCT Contributor
- Joined
- Jan 17, 2017
- Messages
- 13,931
- Reaction score
- 23,234
Thanks @KenAllen15
I'll check the cliff notes and see what I can do to clarify that section ..
notes: ( putting some notes here for now .. )
From Dahua's spec sheet Detect (8 ppf)
Observe (19 ppf)
Recognize (38 ppf)
Identify (76 ppf)
"2. The DORI distance is a measure of the general proximity for a specific classification to help pinpoint the right camera for your needs. The DORI distance is calculated based on sensor specifications and lab test results according to EN 62676-4 , the standard that defines the criteria for the Detect, Observe, Recognize and Identify classifications."
Pixel Density, PPM and PPF in Video Surveillance ( jvsg.com cctv design software )
The DORI standard (based on the IEC EN62676-4: 2015 International Standard) defines different levels of detail for Detection (25PPM), Observation (62PPM), Recognition (125PPM), and Identification (250PPM) for visible light surveillance cameras
DORI (Detection, Observation, Recognition, Identification) How To Measure How Far a Surveillance Camera Can See?
The DORI (Detection, Observation, Recognition, Identification) standard is based on the IEC EN62676-4: 2015 International Standard. Unlike DRI (also known as the Johnson Criteria) for thermal cameras, DORI is designed for visible light surveillance cameras and uses specific measurements in PPM...
www.infinitioptics.com
Discussion at IPVM
Pixels Per Foot, DORI And Large Government Client Who Wants To Standardize On A Particular Camera - IPVM Discussions
IPVM is the world's authority on physical security technology, profiled by Time, The Atlantic, Wired and collaborated with the BBC, NY Times, Reuters, WaPo, WSJ, and more.
ipvm.com
ppf
BSIA ( British Security Industry Association ) = 76 Hanwha = 80 Bosch = 49
more notes: ( again, collecting the references and data so we can update the cliff notes )
What IP Camera Resolution Do You Need?
What IP Camera Resolution Do You Need?
How to determine the resolution required for your IP camera systemBy Bob Mesnik How much resolution do you need? It depends on what you need to see. Do you want to identify who stole your car, or just see that there’s a dog in your yard again. It’s very important to first determine exactly what...
kintronics.com
"Samsung and Axis both say that you need about 40 pixels across the face to identify a person you know. Axis also says that you need up to 80 pixels across the face in challenging conditions. Samsung estimates that you need 40 pixels across the face to be able to see more forensic details such as scars, birth marks and color of eyes. Their IP cameras provide clearer video.
..
In summary, the required pixel resolution is:
Identification of a person that you know – 80 pixels/ft.
Forensic identification of a person you don’t know – 160 pixels/ft.
Best identification when there are poor conditions – 180 pixels/ft.
"
the above refers to
Axis paper ( chart posted earlier )
About Pixel Densities And What They Mean
About Pixel Densities And What They Mean
By Vlado Damjanovski An Internet Protocol (IP) surveillance system is most often used to observe and protect people, objects and people’s activity inside and outside the objects, traffic and vehicl…
www.securitysolutionsmedia.com
AI may require a different number of pixels
Avigilon Face Recognition match confidence:
High: min pixel on face = 80 || Pixels Per Meter = 457 || Pixels per foot = 137
Med: minimum pixel on face = 50-60 || ppm = 286-343 || ppf = 86-103
ref
Camera Guidelines for Face Recognition
help.avigilon.com
IPVM article: Pixels Determine Potential, Not Quality
Pixels Determine Potential, Not Quality
A damaging misconception in surveillance is that pixels = resolution = quality. We explain why it means potential quality and what the limitations are.
ipvm.com
This article has 50-60 pixels per foot as required for identifying
Designing Video Surveillance Systems: Managing Owner’s Expectations |
www.newcomb-boyd.com
Verkada paper: ppf 75+ recommended for people recognition, ppf 150+ for face search and person of interest alerts. Less than 75 ppf => not optimal for people analytics
Bosch Security currently calculates DORI with the following:
DORI (Detect, Observe, Recognize, Identify) is a standard system (EN-IEC 62676-4) for defining the ability of a camera to distinguish persons or objects within a covered area.
- Detect: 25 pixels per meter
- Observe: 63 pixels per meter
- Recognize: 125 pixels per meter
- Identify: 250 pixels per meter
Calculating What You Can See With Your IP Camera
Gets into the math:
Calculating What You Can See With Your IP Camera
What is the right lens and resolution for your IP camera? When you put together your IP camera system, you want to make sure that the camera you select for each location meets your expectations. It is important to first know the objectives for each area you are viewing. Do you want to identify...
kintronics.com
DORI - John Johnson's criteria ... 1957-58 US Army scientist .. this is the start .. we're now in digital domain so naturally things have changed since then
Johnson's criteria - Wikipedia
en.wikipedia.org
Last edited:
Nonamesleft
n3wb
- Joined
- Aug 5, 2016
- Messages
- 20
- Reaction score
- 8
He makes money from affiliate links. If can't make money, he wont make the video.
Mark_M
Getting comfortable
I come from the animated Christmas lighting hobby.
That video sponsor, Holiday Coro.
Nicknamed Holiday Copy because the person behind it sold a certain type of pixel controller for a while, then made a direct clone and sold it for more. His Business is just about up selling and his controllers software is shite.
Absolute terrible things to troubleshoot (I have first hand experience helping others).
Of course; The Hookup just sees money, like money in 'reviewing' Reocrap.
Mark_M
Getting comfortable
Can I also ask what the heck is going on about the consumer market?
I seem to remember a report from IPVM around 2018 mentioned that Ring makes more than Axis. (I'm sure @john-ipvm can fill in.)
So if there's sooo much profit to be made in consumers stuff, why hasn't Dahua/Hikvison stepped up their game in software?
To another point; consumers brands like Amcrest and other OEMs not making 'decent' apps that home owners like.
It's all about the app to home users.
I seem to remember a report from IPVM around 2018 mentioned that Ring makes more than Axis. (I'm sure @john-ipvm can fill in.)
So if there's sooo much profit to be made in consumers stuff, why hasn't Dahua/Hikvison stepped up their game in software?
To another point; consumers brands like Amcrest and other OEMs not making 'decent' apps that home owners like.
It's all about the app to home users.
His testing is flawed at night.
He's testing all the cams at the same time and god knows what else he has attached to his soffits and pointing at the same location.
All the cams have their IR turned on at the same time.
it's not at all representative of what a single cam can do as each cam is getting at least 3 cams worth of IR.
9mins17 for proof.
To be fair it's the same crap testing you get from manufacturers that claim to be going 50m on 3x LED's but they dont do their testing in the boondocks, they do it in downtown Hangzhou with street lighting.
He's testing all the cams at the same time and god knows what else he has attached to his soffits and pointing at the same location.
All the cams have their IR turned on at the same time.
it's not at all representative of what a single cam can do as each cam is getting at least 3 cams worth of IR.
9mins17 for proof.
To be fair it's the same crap testing you get from manufacturers that claim to be going 50m on 3x LED's but they dont do their testing in the boondocks, they do it in downtown Hangzhou with street lighting.
mat200
IPCT Contributor
- Joined
- Jan 17, 2017
- Messages
- 13,931
- Reaction score
- 23,234
Good eye @cctv-dave
During the night he is running the cameras concurrently.
If each camera has their IR lights on, .. then we are indeed not seeing properly compared results, and the scene will have a lot of IR lighting if all the cameras are running their default settings for IR low light image capture.
You see here the subject is in the exact same pose ... thus this is not a repeated test path and any IR lights from all the cameras that are up will provide more IR light to the scene.
You can test cameras concurrently if the cameras are not emitting additional light to your scene ( this is typically the case for daylight .. at night cameras are often adding IR light / or even now "white" light to the scene )
When the subject captured is running, all the cameras are being concurrently tested.
Left Coast Geek
Getting comfortable
A few random comments ...
I have a couple cheap ($49) Reolinks, I thnk they are the RLC-410? they work fine with BI, but their IR illuminator is just way too bright and is a total spider magnet, and yeah, the night vision is not very good for motion, and if I turn down the illuminator the night vision is just pathetic. they work fine in the day but so does everything.
I note they compare a 3.6mm 1/1.8" camera with a 4mm 1/2.5" camera, and thats totally unfair. a 1/2.5" sensor is 0.4", while a 1/1.8" is 0.55", so the equiv of that 4mm 1/2.5" is a 5.5mm lens (round that to 6mm) with the 1/1.8" sensor to get a similar total FOV (4mm / 0.4) * 0.55 == 5.5 mm. I do wish the photography world had a better measure for field of view than mm focal length as thats only comparable across the same size sensor. a 50mm lens on a 35mm camera (24x36mm sensor) is a 30mm lens on an APS-C camera (14x22mm), or a 10 mm on a 1/1.8" (5.3x7.1mm)
(note I round off most optical angles and magnifications to 5 or 10% because the base numbers are nearly always approximate)
I have a couple cheap ($49) Reolinks, I thnk they are the RLC-410? they work fine with BI, but their IR illuminator is just way too bright and is a total spider magnet, and yeah, the night vision is not very good for motion, and if I turn down the illuminator the night vision is just pathetic. they work fine in the day but so does everything.
I note they compare a 3.6mm 1/1.8" camera with a 4mm 1/2.5" camera, and thats totally unfair. a 1/2.5" sensor is 0.4", while a 1/1.8" is 0.55", so the equiv of that 4mm 1/2.5" is a 5.5mm lens (round that to 6mm) with the 1/1.8" sensor to get a similar total FOV (4mm / 0.4) * 0.55 == 5.5 mm. I do wish the photography world had a better measure for field of view than mm focal length as thats only comparable across the same size sensor. a 50mm lens on a 35mm camera (24x36mm sensor) is a 30mm lens on an APS-C camera (14x22mm), or a 10 mm on a 1/1.8" (5.3x7.1mm)
(note I round off most optical angles and magnifications to 5 or 10% because the base numbers are nearly always approximate)
user8963
Known around here
dont let mat or wittaj hear that...they are not to bright. they are unable to control the image or dim the leds ... so everything is just overexposed which is near the camera
never heard that dahua uses anti spider IR leds.. but ok