Saving a cv2.calcHist into a single cell in a CSV file using python
Question:
I am working on a project using python where I need to save properties of a human that is being detected such as clothes, clothes colour and the Histogram of the detection. I wish to save each property into a single cell so that each row represents a single detection.
Now I wish that someone will tell me how it is best to save a histogram array into a single cell.
Is this possible?
Answers:
You need to decide on a format such that it can be read back in easily. The following approach first creates some random histogram data for 3 rows of data. It first writes a header. Next it converts each histogram into text using the np.array2string() function. It uses a ; separator so as not to confuse it with the standard , for the other values. It then removes all unneeded spaces and newlines so the resulting two dimensional array is on a single line. It then uses a standard csv.writer() to write the rows.
I then show how the data could be read back in again and converted back to numpy arrays:
import numpy as np
import csv
import ast
import sys
# Create some data and histograms to save
row = [(0, "Tops", "maroon"), (1, "Socks", "blue"), (2, "Shirts", "white")]
hists = [np.random.randint(0, 20, size=(3, 10), dtype=np.int32) for _ in range(3)]
with open('output.csv', 'w', newline='') as f_output:
csv_output = csv.writer(f_output)
csv_output.writerow(['index', 'type', 'colour', 'hist'])
for item, hist in zip(row, hists):
hist_text = np.array2string(hist, separator=';', threshold=sys.maxsize).replace(' ','').replace('n', '')
csv_output.writerow([*item, hist_text])
# To read it back from the CSV into a numpy array
with open('output.csv') as f_input:
csv_input = csv.reader(f_input)
header = next(csv_input)
for index, type, colour, hist_text in csv_input:
print(index, type, colour)
hist = np.array(ast.literal_eval(hist_text.replace(';', ',')), dtype=np.int32)
print(hist)
So this would give you output.csv containing something like:
index,type,colour,hist
0,Tops,maroon,[[12;6;9;10;2;7;10;8;0;4];[16;10;5;16;15;17;7;14;2;4];[18;17;11;10;19;15;1;1;2;8]]
1,Socks,blue,[[15;5;4;7;18;12;9;5;6;14];[11;6;12;1;2;11;11;15;0;12];[9;11;7;2;14;18;8;13;15;17]]
2,Shirts,white,[[11;17;12;9;2;13;19;5;10;14];[16;16;3;8;5;8;6;0;18;15];[5;7;2;4;16;5;11;17;9;19]]
And whilst reading it back in, it would display:
0 Tops maroon
[[12 6 9 10 2 7 10 8 0 4]
[16 10 5 16 15 17 7 14 2 4]
[18 17 11 10 19 15 1 1 2 8]]
1 Socks blue
[[15 5 4 7 18 12 9 5 6 14]
[11 6 12 1 2 11 11 15 0 12]
[ 9 11 7 2 14 18 8 13 15 17]]
2 Shirts white
[[11 17 12 9 2 13 19 5 10 14]
[16 16 3 8 5 8 6 0 18 15]
[ 5 7 2 4 16 5 11 17 9 19]]
I am working on a project using python where I need to save properties of a human that is being detected such as clothes, clothes colour and the Histogram of the detection. I wish to save each property into a single cell so that each row represents a single detection.
Now I wish that someone will tell me how it is best to save a histogram array into a single cell.
Is this possible?
You need to decide on a format such that it can be read back in easily. The following approach first creates some random histogram data for 3 rows of data. It first writes a header. Next it converts each histogram into text using the np.array2string() function. It uses a ; separator so as not to confuse it with the standard , for the other values. It then removes all unneeded spaces and newlines so the resulting two dimensional array is on a single line. It then uses a standard csv.writer() to write the rows.
I then show how the data could be read back in again and converted back to numpy arrays:
import numpy as np
import csv
import ast
import sys
# Create some data and histograms to save
row = [(0, "Tops", "maroon"), (1, "Socks", "blue"), (2, "Shirts", "white")]
hists = [np.random.randint(0, 20, size=(3, 10), dtype=np.int32) for _ in range(3)]
with open('output.csv', 'w', newline='') as f_output:
csv_output = csv.writer(f_output)
csv_output.writerow(['index', 'type', 'colour', 'hist'])
for item, hist in zip(row, hists):
hist_text = np.array2string(hist, separator=';', threshold=sys.maxsize).replace(' ','').replace('n', '')
csv_output.writerow([*item, hist_text])
# To read it back from the CSV into a numpy array
with open('output.csv') as f_input:
csv_input = csv.reader(f_input)
header = next(csv_input)
for index, type, colour, hist_text in csv_input:
print(index, type, colour)
hist = np.array(ast.literal_eval(hist_text.replace(';', ',')), dtype=np.int32)
print(hist)
So this would give you output.csv containing something like:
index,type,colour,hist
0,Tops,maroon,[[12;6;9;10;2;7;10;8;0;4];[16;10;5;16;15;17;7;14;2;4];[18;17;11;10;19;15;1;1;2;8]]
1,Socks,blue,[[15;5;4;7;18;12;9;5;6;14];[11;6;12;1;2;11;11;15;0;12];[9;11;7;2;14;18;8;13;15;17]]
2,Shirts,white,[[11;17;12;9;2;13;19;5;10;14];[16;16;3;8;5;8;6;0;18;15];[5;7;2;4;16;5;11;17;9;19]]
And whilst reading it back in, it would display:
0 Tops maroon
[[12 6 9 10 2 7 10 8 0 4]
[16 10 5 16 15 17 7 14 2 4]
[18 17 11 10 19 15 1 1 2 8]]
1 Socks blue
[[15 5 4 7 18 12 9 5 6 14]
[11 6 12 1 2 11 11 15 0 12]
[ 9 11 7 2 14 18 8 13 15 17]]
2 Shirts white
[[11 17 12 9 2 13 19 5 10 14]
[16 16 3 8 5 8 6 0 18 15]
[ 5 7 2 4 16 5 11 17 9 19]]