Alexander Zagajewski

Slow update times in crowded images

2

**Describe the bug** A noticable slowdown is experienced when the total number of labels per image exceeds roughly 100. This effectively limits the total number of labels that can be...

Just-in-time file IO and augmentation

1

Hi, Great work! I've noticed in the API the train methods for both the cellpose and size estimation models force the user to load the entire training set into memory...

solve_trig sometimes returns an incorrect root

6

solve_trig by default returns the first root (k=0) of the trigonometric solution, which is sometimes not the correct root. This leads to errors in calculation of x_c which propagates to...

I've noticed the radial distribution is occasionally not calculated correctly, especially when the cell is approaching vertical orientation. This is most evident when trying to reconstruct the image. Tracked down...

Incremental does not handle dask arrays of ndim>2 in estimator training

2

**Describe the issue**: I've been trying to use dask-ml to train large models with multidimensional inputs using Incremental to sequentially pass chunks of the dask array for training. Unfortunately, it...

Multidimensional data inputs

1

I've been trying to train a Convolutional Neural Network on a large, out of memory dataset, using Scikeras as a bridge between Keras and Dask. From some experiments, it seems...

0.32.0 Windows binaries

6

Hi! Any chance we could get Windows binaries on PyPi for the latest release? Pip only finds Windows binaries for 0.31.0, which is not fully compatible with the latest version...