Tool for analysing general embedding 1024 dimension vector

[Markus]

Hi,

I'm testing general embedding model, and my goal is to detect anomalies in series of photos.
General embedding model model returns 1024 dimensional vector with values.

Let's say I run the model with 10 really similar photos to get average values of dimensions.
Then I have the average value of each dimension (1024 values).

After that I do testing with invididual additional photos to find if photo "enough" different or not.

What kind of tool or approach would you suggest for this kind of analysis. I can do some Python and Tensorflow, but I'm wondering would excel be the easiest to visualize and detect the anomalies or is there some other tools you would suggest?

Br,

Markus.

[jared]

Hey @Markus! Spoke to our Research Team about this and they said that you can use l2 distance and tune a threshold for what should be “anomalous” via this site:

https://docs.scipy.org/doc/numpy-1.13.0/reference/generated/numpy.linalg.norm.html

We also have a bit of sample code here. It's not directly for anomaly detection but the same principle holds.

[Markus]

Great! Thanks for help.

[jared]