Navigation Menu

Skip to content

Instantly share code, notes, and snippets.

View svecon's full-sized avatar

Ondřej Švec svecon

View GitHub Profile
@svecon
svecon / download_multiple_images.py
Created May 29, 2018 19:48
Download multiple images using Python
import requests
def DownloadImage(pic_url_prefix, pic_name):
with open(pic_name, 'wb') as handle:
response = requests.get(pic_url_prefix+pic_name, stream=True)
if not response.ok:
print(pic_url_prefix+pic_name, response)
for block in response.iter_content(1024):
@svecon
svecon / keras_cut_network.py
Created August 29, 2017 12:24
Remove last layers of Keras NN model
def remove_layers(model, n)
for _ in range(n):
model.pop()
model.outputs = [model.layers[-1].output]
model.layers[-1].outbound_nodes = []
model.compile(loss='binary_crossentropy', optimizer='adam')
return model
# Example
new_model = remove_layers(model.layers_by_depth[1][0])
@svecon
svecon / receptive_field.py
Last active November 13, 2020 07:52
Calculate receptive field and strides for multiple convolutions
# https://medium.com/@nikasa1889/a-guide-to-receptive-field-arithmetic-for-convolutional-neural-networks-e0f514068807#1780
import math
def receptive_field(n, layers):
"""
Parameters
----------
n: int, input size
layers: array of triplets: (k,s,p)
k: kernel size
s: stride
@svecon
svecon / iPython_notebook_styles.py
Created August 29, 2017 12:02
Make the notebook full width and align output on the right
from IPython.core.display import display, HTML
display(HTML("<style>.container { width:100% !important; }</style>"))
display(HTML("<style> div.cell {display: -webkit-box; -webkit-box-orient: horizontal;-webkit-box-align: stretch;\
display: -moz-box;-moz-box-orient: horizontal;-moz-box-align: stretch;\
display: box;box-orient: horizontal;box-align: stretch;}</style>"))
display(HTML("<style> div.input{min-width:50%;}</style>"))
display(HTML("<style> div.code_cell{width: 70%;float: left;}"
+"div.text_cell{width: 30%;float: right;}"
+"div.text_cell div.prompt {display: none;}</style>"))
@svecon
svecon / logical_puzzles.txt
Created August 28, 2017 12:18
Logical puzzles
A mathematician and a lady strike conversation during a flight. The lady asks the mathematician to guess the age of her three daughters. Mathematician, being unable to answer the questions, asks for some hint(s).
The lady replies: The product of their ages is 36.
Still unable to answer, mathematician asks for another hint.
Hint 2: The sum of their ages is equal to your (i.e. mathematician's) seat number.
Mathematician still cann't answer the question.
Final hint: My (i.e. the lady's) youngest daughter has blue eyes.
Listening to the final hint, mathematician is able to answer the question.
x*y*z = 36 = 2*2*3*3
@svecon
svecon / Bellet-2014-SML-notes.txt
Created August 28, 2017 12:15
Notes for Bellet 2014 survey
Mahalanobis distance dM(x,x′)^2 = (x − x′)^T M (x − x′)
where M is cone of symmetric PSD d*d matrix
Must-link / cannot-link constraints (sometimes called positive / negative pairs):
S = {(xi, xj) : xi and xj should be similar},
D = {(xi, xj) : xi and xj should be dissimilar}.
Relative constraints (sometimes called training triplets):
R = {(xi, xj , xk) : xi should be more similar to xj than to xk}.
Optimization problem that has the following general form: min M { ℓ(M, S,D,R) + λR(M) }
import tensorflow as tf
class EvoNetwork:
def __init__(self, input_size, weights, use_biases=False):
with tf.Graph().as_default():
self.inputs = tf.placeholder(tf.float32, [None, input_size])
hidden_layer = self.inputs
for i, weight in enumerate(weights):
with tf.variable_scope('hidden_layer_{}'.format(i)):
@svecon
svecon / dateToJND.py
Created December 12, 2015 15:01
Converting date to number of days in Julian Calendar
# Converting date to number of days
# Useful when you need to add few days and get a date
# https://en.wikipedia.org/wiki/Julian_day
def toJND(day,month,year):
a = (14-month)//12
y = year + 4800 - a
m = month + 12*a - 3
return day + (153*m+2)//5 + 365*y + y//4 - 32083
@svecon
svecon / ntoskrnl.exe_registry-fix
Last active September 11, 2015 00:53
High RAM Usage of Windows 10 System (ntoskrnl.exe) Process
[HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Services\TimeBroker]
"Start"=dword:00000003
[HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Services\SysMain]
"DisplayName"="Superfetch"
"Start"=dword:00000003
(function (doc) {
var xhr = new XMLHttpRequest();
xhr.onload = function () {
var div = doc.createElement('div');
div.innerHTML = this.responseText;
// div.style.display = 'none'; -- causes gradients not to render!
div.style.visibility = 'hidden';
div.style.position = 'absolute';
div.style.height = 0;
div.style.width = 0;