search

skywolf829 / GSTK

Gaussian Splatting toolkit application. One stop shop for preprocessing your dataset, training your model with human-in-the-loop training, and editing saved GSplat PLY files.
MIT License
4 stars 1 forks source link

Add training statistics view in frontend #22

Open skywolf829 opened 7 months ago

skywolf829 commented 7 months ago

While training, the backend Trainer sends updates to the frontend every X iterations (currently 100). It sends the iteration, maximum iteration, loss, and moving average loss (all from the most recent step). The image from the last iteration is also available but currently not sent (see backend.py on_train_step(...))

In the TrainerWindow on the frontend, a graph showing the training error may be useful. Could be the moving average and just send one update every 100 iterations or so, or maybe you can keep a running list of errors in the backend from each step and send them all at once every 100 iterations. dpg.add_line_series might be good for line chart. Other charts seem to exist like candle charts and histograms.

skywolf829 commented 6 months ago

Moved to web-based, some stuff above is irrelevant now. Backend sends training data updates once a second (loss, milliseconds for train step). Might still be useful to visualize in some bars in the Trainer window.