prabindh

// Call like below
// boxedMat will contain resized image
// boxedMat = __LetterBoxResize(floatMat, net->w, net->h);
//
// Taken mostly from https://jdhao.github.io/2017/11/06/resize-image-to-square-with-padding/

cv::Mat ArapahoV2::__LetterBoxResize(cv::Mat img, int w, int h)
{
    cv::Mat intermediateImg, outputImg;
    int delta_w, delta_h, top, left, bottom, right;

prabindh

# -*- coding: utf-8 -*-
"""
Created on Mon Jun 13 16:59:53 2016
From - https://stats.stackexchange.com/questions/218407/encoding-angle-data-for-neural-network
@author: Ari
"""

from numpy import savetxt, loadtxt, round, zeros, sin, cos, arctan2, clip, pi, tanh, exp, arange, dot, outer, array, shape, zeros_like, reshape, mean, median, max, min
from numpy.random import rand, shuffle
import matplotlib.pyplot as plt

prabindh

import os
import json

def path_to_dict(path):
    d = {'name': os.path.basename(path)}
    if os.path.isdir(path):
        d['type'] = "directory"
        d['children'] = [path_to_dict(os.path.join(path,x)) for x in os.listdir\
(path)]
    else:

prabindh

# https://stackoverflow.com/questions/56069411
# The TF code
score_inputs = tf.placeholder(np.float32, shape=(None, 100))
targets = tf.placeholder(np.float32, shape=(None), name="targets")

l2 = tf.contrib.layers.l2_regularizer(0.01)

first_layer = tf.layers.dense(score_inputs, 100, activation=tf.nn.relu, kernel_regularizer=l2)
outputs = tf.layers.dense(first_layer, 1, activation = None, kernel_regularizer=l2)

prabindh

from keras.models import Sequential
from keras.layers import Dense
import numpy as np

def normalize_angles(phases):
    phases = phases + np.pi
    phases /= (2 * np.pi)
    return phases

def build_fourier_mnist():

prabindh

CuDNNGRU stateful implementation, TensorFlow backend 
Layers: (300,80,150) -- (encoder, latent, decoder) -- (selu, tanh, tanh)
return_sequences=True for all layers
GaussianNoise + Dropout at (=after) input, AlphaDropout at encoder, Dropout at latent
BatchNormalization between encoder and latent, latent and decoder
Output: TimeDistributed(Dense(units=input_dim, activation='linear'))
batch_size=25, timesteps=400, input_dim=16 - 25 separate, 10-min sequences fed 400 timesteps (=1 sec) at a time (as 10*60=600 'windows' in parallel, non-shuffled)
reset_states() applied before testing on new x25 10-min sequences
model.fit, or train_on_batch for training

prabindh

from keras.datasets import mnist 
from keras.layers import Dense, Input, concatenate,subtract, Lambda
from keras.losses import binary_crossentropy
from keras.optimizers import SGD

(train_x, train_y), (test_x, test_y) = mnist.load_data()
train_x = (train_x / 255.0).reshape(-1, 28*28) 
test_x = (test_x / 255.0).reshape(-1, 28*28)

inp1 = Input(shape=(28*28,))

prabindh

import numpy as np
from numpy.lib.stride_tricks import as_strided
import tensorflow as tf
import time

def conv2dTrickster(a, b):
    a = as_strided(a,(len(a),a.shape[1]-len(b)+1,a.shape[2]-b.shape[1]+1,len(b),b.shape[1],a.shape[3]),a.strides[:3]+a.strides[1:])
    return np.einsum('abcijk,ijkd', a, b[::-1,::-1])
    
def conv2dSimple(image, filter):

prabindh

# conv1d for timeseries spike
model = models.Sequential()

model.add(layers.Conv1D(filters = 1,
                        kernel_size = 10,
                        activation = 'relu',
                        input_shape=(timesteps,1)))
model.add(layers.GlobalMaxPooling1D())
model.add(layers.Flatten())
model.add(layers.Dense(1, activation = 'sigmoid'))

prabindh

$ diff ~/Downloads/yolov3.cfg-master.txt ~/Downloads/yolov3.cfg-alexeyab.txt
3,4c3,4
< # batch=1
< # subdivisions=1
---
> batch=1
> subdivisions=1
6,9c6,9
< batch=64
< subdivisions=16