William Lynn

icon indicating an email [email protected]

icon location Minneapolis MN icon location C, Java, Kotlin, Android, KMP, AI, LLM and large datasets.

Results 7 comments of William Lynn

The difference between v2 and v3 version

I am also having the same issue, V3 always gives the WRONG bounding boxes

Not getting proper rect

I have the same issue, I am resizing the input image to 320 x 320. and it finds the correct number of faces (34), BUT the bounding boxes are wrong....

Cannot find dependencies

this repo should have an option to work Without JavaFX, because some people just want to use the algorithms

Different probability predictions everytime

I this is because the matching of faces to see if face A is face B needs some work. I've tried this library for months and the results are always...

How can we retrain the model on custom dataset (Indian Faces)?

> I would like to re-train the model on custom data-set. How can we achieve that? Can you share the source indian dataset of faces ?

Increase the number of landmarks detected by MTCNN

This would require a complete retaining of the network I think.

Regarding the MIT License of the models

I agree with you, if the source images for training can't be used for commercial use. Any work derived from those images can't be used for commercial work either. I...