| T O P I C R E V I E W |
| cpstevenc |
Posted - Jan 30 2025 : 00:54:02
Looking more into IEVision... and what it can maybe do for us.
I am sitting on maybe near 100 TB of unique mostly categorized vehicle images.... maybe more.
I would like to some how use these images in a way to train something.
The training would be for say... show a picture of a car from most angles... and I could tell you what make and model it is.
Probably not year unless its a one off model year or something silly.
Or, I could tell you that this picture is vehicle wheel, and that image of the wheel is used this year/make/model...
Or that what I am looking at, is a fender. Or a center console. A Strut. Doesn't need to know really what make/model/mfg/ect/ect it is.. just be able to know what kind of car part I am looking at in the picture. So if you give me a picture of a hood, but tell me it is a front seat. I could flag that as a big nope, that is a hood you are showing me.
Stuff like that.
I tried the Classifier training demo, as it seemed to be kind of vehicle based on data.
But not sure it works well or not or if it can do close to what I am even thinking here...
Lots of things pick up on in it that are not vehicles, such as side of buildings.
Maybe due to quality of images? maybe due to LBP vs HAAR? I haven't tried HAAR yet as even on a 5950X to run this data is taking forever to do while I write this to run 15 stages. Doesn't seem to be using more than one core, so single threaded? If so could that be improved to run on X cores to mass process stuff quicker?
Or maybe I am looking at the wrong things here over all and over my head?
Basically trying to find out if anyway I can use ImageEN / IEVision to leverage 100TB of images to train "something" in a way that can be used to detect what an image of a vehicle or vehicle part is... or just what the part COULD be.
Or this a case of "just use some silly python script project out there" as this is way out of the scope of ImageEN/IEvision?
|
| 3 L A T E S T R E P L I E S (Newest First) |
| xequte |
Posted - Feb 02 2025 : 22:29:21
Hi
Yes, we can confirm there is an issue with OpenCL with some systems. Can you confirm your video card, and whether OpenCL.dll or cl.dll exist in C:\Windows\System32\
Nigel
Xequte Software
www.imageen.com
|
| cpstevenc |
Posted - Jan 30 2025 : 21:44:02
I'll have to dig more into this.
Using the Neural Net Processing demo, using the GPU option the program just closes.
Not sure if i need some other DLL's for this to work? didn't see anything mentioned in
the readme.txt
Then I guess I need to read up and learn how to build model data / prototxt and all that.
|
| xequte |
Posted - Jan 30 2025 : 13:15:36
Yes, that is certainly what it is trained for. You should take a look at the OpenCV documentation for it:
https://docs.opencv.org/4.x/db/d28/tutorial_cascade_classifier.html
That said, you are probably better to use a neural net solution for it, these days:
http://www.imageen.com/help/TIEVisionNNet.classifyObject.html
https://www.google.com/search?q=opencv+neural+net+classify+image
Nigel
Xequte Software
www.imageen.com
|
|
|
Search
