def exist_digit_data(t_file):
#checks to see if digit_data.csv exist. It hold the classification
# labels and the vectors. It is also when counting the number
# of each individual class.
t_prefix=set_file_name_prefix(0)[0]
from os import path
t_file_name=set_file_name_prefix(0)[0]+t_file
#print(t_file_name)
return int(path.exists(t_file_name))
def save(t_canvas_image):
#function used to save imaages to a csc file.
#It takes in the full canvas. It separates the grid squares
# (via separate_digits), refromats them (via reformat_mnist) and
# reshapes then into a one dimensional vector that is written to
# a csv file that can be used as input to different machine
# learning programs. The output includes a class label. The label
# has the same value for all the digits on the filled out grid sqaures.
# Only one class at a time can be entered for each set of grid squares
import pandas as pd
import cv2
import numpy as np
from digit_capture import get_global_settings as gs
d_2=gs('d_2')
grid_count=gs('grid_count')
t_dest=set_file_name_prefix(0)
fname_pfx=t_dest[0]
t_path=t_dest[1]
t_class=gs('t_class')
final_side_size=gs('final_side_size')
df_class=make_class_list(t_class,(grid_count*grid_count))
df_column_names=make_col_list((final_side_size*final_side_size))
out_array = np.zeros(((grid_count*grid_count),(d_2*d_2)),dtype='uint8')
(t_digits,t_reformat_dg)=separate_digits(t_canvas_image,d_2,grid_count)
new_data=reformat_mnist(t_reformat_dg,0)
df_dta=pd.DataFrame(data=new_data)
df_dta=pd.concat([df_class,df_dta],axis=1)
write_digit_data(df_dta,fname_pfx+'digit_data.csv')
def get_dict_from_file(t_file):
#reads in a dataframe and converts it into a dictionary.
#The dictioary holds a lot of values of global variables.
from digit_capture import set_file_name_prefix
import pandas as pd
fname_pfx=set_file_name_prefix(0)[0]
#print(fname_pfx)
df=pd.read_csv(fname_pfx+t_file)
t_name=list(df['Name'])
t_value=list(df['Value'])
t_type=list(df['Type'])
t_value=set_correct_type(t_value,t_type)
res = {t_name[i]: t_value[i] for i in range(len(t_name))}
return res
def get_val_from_dict_csv(t_key):
#input is a key. That is the row entry key in the dictionary_inputs.csv file.
#Read the file into pandas. Convert the Name column into a list. Find the
#index where the lest entry = the_key. Use the index to get the value
#from a list made from th Value column in the file. Return the value as
#a string
import pandas as pd
t_prefix=set_file_name_prefix(in_colab())[0]
t_file=t_prefix+'dictionary_inputs.csv'
df=pd.read_csv(t_file)
t_list_n=list(df['Name'])
t_pos=t_list_n.index(t_key)
t_list_v=list(df['Value'])
return str(t_list_v[t_pos])
def output_data(t_data):
#Outputs the image to a csv file
from digit_capture import get_global_settings as gs
t_dest=set_file_name_prefix(0)
fname_pfx=t_dest[0]
dmy=len(t_data)
import numpy as np
import cv2
tAry=np.asarray(t_data)
#tAry
final_side_size=gs('final_side_size')
grid_count = gs('grid_count')
t_class=gs('t_class')
df_class=make_class_list(t_class,(grid_count*grid_count))
df_column_names=make_col_list((final_side_size*final_side_size))
pxl=gs('pxl')
line_wd=gs('line_wd')
canvas_width = grid_count*pxl
canvas_height = grid_count*pxl
prev_x=-999999;
prev_y=-999999;
x_coords=[]
y_coords=[]
width = canvas_width # canvas width
height = canvas_height # canvas height
center = height/2
white = (255, 255, 255) # canvas back
t_bd1=5
t_bd2=175
d_1=1
d_2=gs('d_2')
d_3=120
d_4=180
wdh=1
#print(type(t_data))
tAry=np.asarray(t_data,dtype="uint8")
#tAry.shape
#tAry2_dim0=int(len(tAry)/4)
#tAry2=np.reshape(tAry,(tAry2_dim0,4))
#nw_image=np.reshape(tAry2[0:tAry2_dim0,0],(canvas_width,canvas_width))
#cv2.imwrite(fname_pfx+'python_version.jpg',nw_image)
#print(' point 3 '+str(nw_image.shape))
#print(type(nw_image[0,0]))
nw_image=np.reshape(tAry,(canvas_width,canvas_width))
save(nw_image)
def get_class_counts(t_file):
import pandas as pd
#gets a dictionary with the counts of classes that are in
#digit_data.csv. If the file does not exist, it returns a dictioary
#with only a single entry with 0 for count. This is how the
#situation where no digit_data.csv exists is handled. Groupby
#returns a dictionary
dd={'replace_dict': {'?class02?': 0}}
fname_pfx=set_file_name_prefix(0)[0]
#print(fname_pfx)
if(exist_digit_data(t_file)==1):
df=pd.read_csv(fname_pfx+t_file,header=None)
#print(df.shape)
t_df=df.iloc[:,0:2]
t_list=['class','replace_dict']
t_df.columns = t_list
dd=t_df.groupby('class').count().to_dict()
return dd
def insert_mods_to_js_file(t_js_file,the_dict):
#This function takes as input a js file and a dict.
#The js file has wild card values for certain varibles
#And functions. These wildcards get swapped for values
#cross referenced in the dictionary that is also input.
#This allows the python to communicate with the javascript.
from digit_capture import set_file_name_prefix,in_colab
from digit_capture import read_pickle,write_pickle
fname_pfx=set_file_name_prefix(0)[0]
#print(fname_pfx)
#pk_file=t_js_file[0,-3]+'.pk'
#the_dict=read_pickle(fname_pfx+pk_file)
#file1 = open(fname_pfx+t_js_file,"r+")
file1 = open(fname_pfx+t_js_file,"r+")
the_js_data=file1.read()
for old,new in the_dict.items():
the_js_data=the_js_data.replace(old,str(new))
return the_js_data
def update_class_number(nw_class_index):
#reads in dictionary_inputs.csv which holds values for global
#variable where the name is the global variable name. Once read in,
#it updates class label. This needs to be done via user so it can be changed
#as a user enters different digit classes.
import pandas as pd
t_list02=[]
for i in range(0,10):
t_class='?class0'+str(i)+'?'
t_list02.append(t_class)
nw_class=t_list02[nw_class_index]
t_prefix=set_file_name_prefix(in_colab())[0]
t_file=t_prefix+'dictionary_inputs.csv'
df=pd.read_csv(t_file)
t_list=list(df['Name'])
t_pos=t_list.index('t_class')
df.at[t_pos, 'Value']=nw_class
t_pos=t_list.index('?selval?')
df.at[t_pos, 'Value']=nw_class_index
df.to_csv(t_file,index=False)
return df