Full Body Tracking and Recognition

Hi there. I am new to Python and OpenCV, so I apologize if my problem is a bit obvious to you.

I google a way to recognize people without using facial recognition. I imagine capturing it with a photo and analyzing the colors so that in a group of people the computer knows who I indicated.

Can anyone help me with my problem? Please. It can be with OpenCV or another way.

Greetings.

kindly look at the person re-identification sample here

2 Likes

Thanks for your answer.

I am looking for some documentation, article or video that has used the code you suggest. Because as I am new I do not understand it at once. I hope it is not too much to ask, but if you know of any I will appreciate it.

you’ll find some references in the original pr

Zheng F, Deng C, Sun X, et al. Pyramidal person re-identification via multi-loss dynamic training[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2019: 8514-8522.

Yu F, Jiang X, Gong Y, et al. Devil’s in the Details: Aligning Visual Clues for Conditional Embedding in Person Re-Identification[J]. arXiv e-prints, 2020: arXiv: 2009.05250.

in the end, there a pretrained cnn, you process images through it , and receive a simple 1d feature vector, that can be easily compared with L2 or cosine norm .

1 Like

I followed the indications that the code places at the beginning (send to this page GitHub - ReID-Team/ReID_extra_testdata).

I downloaded “query”, “galley” and the 3 “.onnx” models it shows (lite, medium, lage). I tried with each model and all of them get an error.

With the “Lite” model I got the following error:

Please don’t judge me, remember that I am new. Tell me if I made a mistake.

please do not post screenshots of code or errors (anywhere on the net !) but TEXT

and your image is empty (wrong path)

1 Like

Sorry.

I write the exact path. What I do see is that it says there are 4 images, when there are only 3.

Feature Extraction for images in /Users/MyUser/Desktop/test/query Batch: 0 / 4
Feature Extraction for images in /Users/MyUser/Desktop/test/gallery Batch: 0 / 21

Correction, it works already. Is that Apple created an invisible file called ".DS_Store"

1 Like

I save an image of Biden as “Query” and of “Gallery” I saved several of Biden and others of Obama (ordered in no order of name). As a result I received this:

What the code does is order people? How do I know if it works?

I honestly don’t know what the code does exactly, I don’t know if it identifies it well or not.

those are the 5 closest matches from the “gallery”, sorted by distance

whether this approach is “a good fit” for your problem
– how would we know ??

can you try to be more specific ?

1 Like

then use tracking

but seriously, it’s YOU, who has to do the research, not me.
“asking around” is no replacement

Thanks for your help, I appreciate it.

My problem is knowing that X person is the one that matches what the ReID says.

I am investigating, what happens is that if I ask you can give me valuable information that I do not know and perhaps ignore in the research, remember that I am new to the subject.

apologies for getting somewhat impatient with you before here …

that will still need to isolate the person from the “rest of the image” (HOGDetector or yolo)
once you have an initial rectangle, you can start a tracker to get cheap updates ob the position

once that fails, you have to get fresh input:

that’s where the reid comes in. it’s just “oppositely winded”. the “gallery” in the sample code are the persons in your image. (and again, you need some object detection code to isolate them, before you can “classify” those)
once it has found you there, you can start another tracker instance

… and so on …

1 Like