ModelView | Object Oriented OpenGL

I have a Three.js project with an animation going on and I would like to find the skeleton bones position at different times.

If I go for example to: https://modelviewer.dev/examples/animation/index.html and find the Three.js scene:

My issue is that I'm trying to make a SurfaceView display a 3D model that's just a small part of the whole application layout. The layout includes multiple views for displaying other data; however, as soon as I create the ModelViewer the SurfaceView vanishes. When I use version 1.9.9 of Filament the layout is displayed, but heavily pixelated and seems to make other layout elements disappear. With v1.7.0 the other elements don't dissapear, but nothing is displayed.

I'm using the following version in Gradle file:

I'm using Flask-Admin to create views of my database tables. One of our tables has a "color" column and it's "editable" meaning if you click on it, you can edit the text directly and it'll update the database with AJAX. It's a text field and I'm expecting hex colors like "#3CA4FF".

I want this "editable" color column to have an HTML5 color picker widget to aid in color selection (i.e. ).

Here's my Python code for:

1. The SQLAlchemy table model class
2. Custom ColorField wtforms field
3. The Flask-Admin model view class

The following works fine to add my custom ColorField to the "edit" form, but what I really want is to edit it using the column_editable_list option.

I have a single text string stored in a SQL table which contains all of the text below. The format is XML but the field definition is varchar.

I am using SQL Server 2012 to query this data:

My ViewModel:

I am writing low level 2D/3D rendering engine as a part of display driver for MCU platform in C++ and I hit a wall with perspective projection in 3D and face culling.

Lets assume my 3D engine is using M,V,P matrices (in the same manner as Model,View and Projection matrices in OpenGL fixed pipeline).

The idea is to convert rasterized face normal into view local coordinates (using MV) and test the sign of coordinate corresponding to view direction. The same can be done with dot product between camera view direction and the normal itself. And according to sign either rasterize or skip face. This works well for parallel rays projections ... however with perspective projections this leads to false positives (you can see faces "visually" tilted away up to some angle).

For filled surfaces this poses only performance hit as depth buffer remedies the artifacts so render looks like it should. However wire frame is a problem:

The remedy is to transform vertexes of face by MVP and do the perspective divide. And then re-compute the normal from the result and use that for face culling:

However that is a lot of more operations which on slow platforms like MCU could pose a performance problem. So my question is:

If it is possible How to use face normal for back face culling safely?

I have tried to transform the normal locally by transforming face center and its small displacement in normal direction by MVP with perspective divide and then recomputing the normal from these 2 points. Still twice the operations then using directly normal but better than 3x. However result was not correct (looked almost identical to using normal directly).

I am thinking of somehow computing the tilt angle for given projection / location and test the:

I am trying to use Google's model viewer to load 3D models. It has a button. But I want that button to be visible after the model is completely loaded. So, I used to use this Vanilla JavaScript code

I am using a flutter library called Model Viewer ModelViewer

That's a very nice plugin which works like charm on my dev version of that app. I am using it like this:

I'm trying to convert world coordinates to screen coordinates. I have available: fov, screen width, screen height, camera position, camera angle and obviously the position of the object in world space.

This is what I tried: