I’ll also be showing the 4MP version in a week or 2. Stay tuned though for a new 4K video as part of the challenge that was set me.
Review: EmpireTech IPC-T58IR-ZE-S3 / Dahua 5842T-ZE-S3 - 8MP / 4K Turret Camera
- Thread starter Wildcat_1
- Start date
Yeah I can see the difference in the AI features as I throw Metadata and Face recognition on basic 2000 series cameras and it works (albeit faces not as good as the 5000 series and was only for testing purposes). And its not as limited as with the EI. But I definitely have more customers picking the EI due to the price.
I wish the XI had Acupick in the web service since for my customers those are rack mounted locked away in IT rooms with no local access to them. Its not something the customer will stick on his office desk. So I havent been able to test the Acupick on that yet.
The only other issue I see with the XI is as I use it over p2p like i do other units (due to the client's IT restrictions/time to open ports etc), it keeps asking to reinstall the webservice. It seems this webservice is based on the IP or ports. And even when the webservice is installed it will also say Plugin Not Compatible, and then load the video. Not sure if thats a bug Dahua know about or not.
Separately, the way the 5000 series Metadata shows the plates, it would be nice to be able to search for that plate. But that seems to be saved in the camera right, therefore requiring the LPR camera?
And another thing I always wondered, why we cant search for the name of a person in the face search.
I name the staff in the database for example but when I do a search it does not find anything.
However the face alerts shows the name.
I must say Face Recognition has improved 100% since i first tested it on the Lite AI NVR back in 2020.
At 80% its giving me faces I can barely recognize myself but somehow it sees the similarity.
DMSS people alerts is great too, although IMO the app itself needs alot of work (crashing all the time for me).
Quick question about the XI as Dahua specs vary. Its basically 8 channels of AI by recorder for everything right? So 4x Face Recognition, 2 Video Metadata, 2 SMD would max it out, or is SMD not counted? On the EI when I did Face Rec on 1 channel it limited me to 4 x SMD by recorder max. As you mentioned the LPR I may want to put that on a couple cameras like the 5442 and or the 5842. I may end up doing 4x Face Rec and 4x LPR. If thats even possible. Using Metadata, Face Detect, and SMD (2000 series) by camera.
No.. XI have 24 or 32 AI points (depending of model, firmware and did You use separate video for HDMI2/3/4)...
SMD/IVS is using 1 AI point, full Face Detection is using 4 AI points, Face Recognition (with Face Detection done on camera) is using 1/2 AI points, Video MetaData is using 4 AI points..
AcuPick 1.0 is using 0.5 AI points...
there is some small variability depending of firmware version and how You join different AI functions (sometimes limits are higher as declared, sometimes lower)..
I received reply from my Dahua rep for my many queries about AcuPick 2.0 firmware for XI line...
they are working on that.. no ETA.. no alpha/beta versions yet.. this is reply from China HQ...
maybe someone else will get a better result..
Yes I’m providing feedback and some bug fixes to their dev team on implementing ASAP with a few ideas since this was targeted for July.
Ye
I’ll be doing a review on the fixed 4MP as well
No its not connected to a monitor yet. I only have the 1 XI as yet. Eventually may get a monitor but all they have is a 90's crt.
Also, Metadata at one of my clients using the IPC-T54IR-ZE (plan to upgrade to the 8MP to get faces and plates double the distance), averages about 1100 humans and 1500 vehicles daily, most before 12 noon. But I can see the need for the 8MP for plates and faces further away, as I have that at another location. I had the 8MP set to default and it was catching the main road about 100 feet away (zoomed in to 60), it had 70 pages in 1 hour full of vehicles from the main road, including mine, and it didnt miss most by the looks of it. Took it off the main road.
The aucpick 2.0 still have some more work to do, i will check with dahua team after their national holiday, Oct.1st-7th is their holiday. People use BI or 3rd party software this function not useful.
-XI firmware as i know they are working on it, but need time to finish them, the XI is more expensive and comes out much later than the -EI products, so that is why the -EI more update.
-XI firmware as i know they are working on it, but need time to finish them, the XI is more expensive and comes out much later than the -EI products, so that is why the -EI more update.
Yep, I’ll be filing more bugs and updates this week too. Busy times but needed for great products 
What is the benefit to MetaData vs IVS? Is Metadata better for busy areas or also good at a residential home? Thanks
IVS is for perimeter protection - to catch humans/vehicles which entered a specific area.
Video MetaData is for building a searchable and annotated database of all the people/cars the camera sees.
It is best to test these differences on your cameras.
Video MetaData is for building a searchable and annotated database of all the people/cars the camera sees.
It is best to test these differences on your cameras.
It is also good practice to implement both based on differing areas of your location and target requirements within those.
All, posting here (will add to original posts to make it easier to find in future) and will post a separate thread too on the challenge I've been alluding to.
Hope everyone had a great weekend. I wanted to post this video I took using the new S3 cams I’ve been testing. Specifically I was challenged via some forum members and YouTube viewers to push these new cameras (4K variant) to the limit and attempt to use them for LPR (NPR) recognition. To be clear this is not to be confused with Automated License / Number Plate Recognition which as I’ve shown before has to be supported by the camera through OCR processing as part of the AI algorithm. These cameras actually DO support that BUT critically this has not been implemented yet in FW (read my bug list in the main review HERE).
I’ve shown before and most people are aware that 2MP, 4MP cams at Z4 or Z12 do a good job in these kinds of situations BUT as I’ve mentioned before, light requirements, noise etc increase dramatically for 8MP/4K, making this a harder task.
So with that said, in this video I want to show you what’s possible, how you can (and MUST) tame the over-sharpening and not only demonstrate the results but share the settings I put in place when undertaking this challenge, including making it more difficult on myself. Unfortunately YouTube adds compression artifacts but I've tried to ensure those are kept to a minimum as always.
In case you're asking, YES, these do great for LPR and even ALPR (using on NVR AI, for the moment), during the day.
Hope this helps, enjoy ! Please do remember to like & subscribe to the video, it does help at least get the content promoted in the algorithm and these videos do take a LOT of time to produce.
Thanks
Do remember to select 4K for playback.
Hope everyone had a great weekend. I wanted to post this video I took using the new S3 cams I’ve been testing. Specifically I was challenged via some forum members and YouTube viewers to push these new cameras (4K variant) to the limit and attempt to use them for LPR (NPR) recognition. To be clear this is not to be confused with Automated License / Number Plate Recognition which as I’ve shown before has to be supported by the camera through OCR processing as part of the AI algorithm. These cameras actually DO support that BUT critically this has not been implemented yet in FW (read my bug list in the main review HERE).
I’ve shown before and most people are aware that 2MP, 4MP cams at Z4 or Z12 do a good job in these kinds of situations BUT as I’ve mentioned before, light requirements, noise etc increase dramatically for 8MP/4K, making this a harder task.
So with that said, in this video I want to show you what’s possible, how you can (and MUST) tame the over-sharpening and not only demonstrate the results but share the settings I put in place when undertaking this challenge, including making it more difficult on myself. Unfortunately YouTube adds compression artifacts but I've tried to ensure those are kept to a minimum as always.
In case you're asking, YES, these do great for LPR and even ALPR (using on NVR AI, for the moment), during the day.
Hope this helps, enjoy ! Please do remember to like & subscribe to the video, it does help at least get the content promoted in the algorithm and these videos do take a LOT of time to produce.
Thanks
Do remember to select 4K for playback.
today I had an epiphany - why does 5442 have the ANPR engine (license plate OCR) partially enabled?
If you have full Video Metadata enabled - 5442 takes snapshots of the license plate. In 95% of cases (including at night) it manages to catch a sharp record of the car plate number..
To select from multiple frames of the video footage, those that have a sharp license plate - the camera must have some mini form of OCR that will determine whether this plate and letters on it is sharp or not..
That is why 5442 has ANPR partially enabled. It accepts the configuration for ANPR and behaves as if it were doing OCR of plates. They used the existing code for ANPR - they simply blocked the final phases (finishing OCR, generating an entry with the plate number).
This is intentional, the 5442 is not supposed to have full ANPR (there is 7442 for that) but they wanted to support plate snapshots for cars (similarly to face snapshots for people).