maweigert

""" 

Plots the decision boundary and the loss as a function of training time/epochs 
for perceptron (no hidden layer) or single hidden layer MLP for difefrent datasets ('moons', 'blobs', 'xor')


requires:

pip install torch torchvision torchaudio matplotlib tqdm seaborn moviepy scipy scikit-learn

maweigert

""" 
Flip a fair coin 100 times—it gives a sequence of heads (H) and tails (T). 
For each HH in the sequence of flips, Alice gets a point; for each HT, Bob does,
so e.g. for the sequence THHHT Alice gets 2 points and Bob gets 1 point. 
Who is most likely to win?


https://x.com/littmath/status/1769044719034647001?s=20

maweigert

# Makefile to create arxiv submission conforming single latex files. Created originaly by Uwe Schmidt 

all: pdf


pdf: main.tex
	pdflatex main.tex
	# http://stackoverflow.com/a/21935690
	-biber main
	pdflatex main.tex

maweigert

name: dl-image-mw
channels:
  - defaults
  - nvidia
dependencies:
  - python=3.9
  - cudatoolkit=11.0.*
  - cudnn=8.0.*
  - jupyter
  - numpy

maweigert

"""
The default momentum of MobileNet (for running average calculation) in tf seems to be very high (0.99?), resulting in huge (and suprising) differences in train vs evaluate loss

https://github.com/keras-team/keras/issues/6977

Setting it to something sensible (e.g. 0.9) fixes this 

https://stackoverflow.com/questions/65415799/fit-works-as-expected-but-then-during-evaluate-model-performs-at-chance/66692755#66692755
"""

maweigert

# Average-Downsampling example in OpenCL (via gputools)
# please install it first via
# pip install gputools

import numpy as np
from gputools import OCLProgram, OCLArray, get_device
from timeit import default_timer
from skimage.transform import downscale_local_mean

# opencl kernel

maweigert

import tensorflow as tf
import numpy as np
import psutil
import os 
from stardist.models import Config3D, StarDist3D

os.environ['CUDA_VISIBLE_DEVICES']= "0"# '1,0'

from tensorflow.keras.utils import Sequence
import stardist

maweigert

/* QuPath-Script to export annotations to label tif images (e.g. to be used for stardist) 

 Use "Run for project" to export annotations for all images within a QuPath project

 Afterwards both images and mask tiffs will be stored in the project subfolder 
 
 ground_truth
 ├── images
 └── masks
 

maweigert

import numpy as np
import matplotlib.pyplot as plt
from scipy.ndimage import label, zoom

from stardist.matching import matching
from stardist.plot import render_label, render_label_pred 
np.random.seed(42)

def create_example():
    rand_gt = np.random.uniform(0,1,(8,8))

maweigert

#@ ImagePlus imp
#@ OUTPUT ImagePlus result

from ij import IJ
from ij.plugin.frame import RoiManager

result = IJ.createImage("Labeling", "16-bit black", imp.getWidth(), imp.getHeight(), 1)
ip = result.getProcessor()
rm = RoiManager.getInstance()