Create a python script to convert an object detection annotations csv file into a json file in the format of a dictionary with keys "images" and "annotations". Value for the "images" key should be a list of entries, one for each image of the form $\{"$file$\_$name": image$\_$name, "height" : height, "width" : width, $"$id$"$ : image$\_$id$\}.$ Value of the "annotations" key should be a list of entries, one for each bounding box of the form $\{"$image$\_$id$"$ : image$\_$id$,$ "bbox" : $[$xmin$,$ ymin, xmax, ymax$],$ "category$\_$id$"$ : bbox$\_$label$\}.$ Our current csv file has the first column of data for the file$\_$name, the second column has the bounding box or "bbox" for X$,$ the third column has the bounding box or "bbox" for Y$,$ the fourth column has the image width, the fifth column has the image height, and the sixth column has the region id or "category$\_$id$".$ Make sure the script is using pandas and or numpy to read in the values as whole numbers without any decimals at all and it verifies there's no decimals being added in or read in from the csv file. Try the script out with a csv file with some integers in columns to make sure the script works as described above.

Here's the code I have so far and it gave me the error of "error converting row $0$: invalid literal for int$()$ with base $10$: $\text{'}55\mathrm{.}0\text{'}$

and in the csv file the value is just $\text{'}55\text{'}$ without the added decimal.

code:

import pandas as pd

import json

from pathlib import Path

def convert$\_$csv$\_$to$\_$json$($csv$\_$file$\_$path, json$\_$file$\_$path$)$:

# Ensure the input and output file paths are Path objects

csv$\_$file$\_$path $=$ Path$($csv$\_$file$\_$path$)$

json$\_$file$\_$path $=$ Path$($json$\_$file$\_$path$)$

# Check if the CSV file exists

if not csv$\_$file$\_$path.is$\_$file$()$:

print$($f$"$Error: The file $\{$csv$\_$file$\_$path$\}$ does not exist!"$)$

return

# Initialize lists for images and annotations

images $=[]$

annotations $=[]$

image$\_$id$\_$map $=\{\}$ # To track and assign unique IDs to each image

image$\_$id$\_$counter $=0$ # Counter for image IDs

# Read the CSV file using pandas

try:

df $=$ pd$.$read$\_$csv$($csv$\_$file$\_$path, dtype$=$str$)$ # Read all columns as strings

except Exception as e:

print$($f$"$Error reading CSV file: $\{$e$\}")$

return

for index, row in df$.$iterrows$()$:

# Ensure that each row has the expected number of columns

if len$($row$)<6$:

print$($f$"$Warning: Row $\{$index$\}$ does not have enough columns. Skipping..."$)$

continue # Optionally skip rows with insufficient columns

# Extract data from the DataFrame row

file$\_$name $=$ row$[0].$strip$()$ # File name as string

# Initialize variables

bbox$\_$x $=$ bbox$\_$y $=$ image$\_$width $=$ image$\_$height $=$ category$\_$id $=$ None

# Extract integer values from the row, converting strings to integers

try:

# Remove any decimal points and convert to int

bbox$\_$x $=$ int$($float$($row$[1].$strip$().$replace$(\text{'}\mathrm{.}0\text{'},\text{'}\text{'})))$ # Bounding box X

bbox$\_$y $=$ int$($float$($row$[2].$strip$().$replace$(\text{'}\mathrm{.}0\text{'},\text{'}\text{'})))$ # Bounding box Y

image$\_$width $=$ int$($float$($row$[3].$strip$().$replace$(\text{'}\mathrm{.}0\text{'},\text{'}\text{'})))$ # Image width

image$\_$height $=$ int$($float$($row$[4].$strip$().$replace$(\text{'}\mathrm{.}0\text{'},\text{'}\text{'})))$ # Image height

category$\_$id $=$ int$($float$($row$[5].$strip$().$replace$(\text{'}\mathrm{.}0\text{'},\text{'}\text{'})))$ # Category ID

except ValueError as e:

print$($f$"$Error converting row $\{$index$\}$: $\{$e$\}.$ Assigning default values $(0$ for bbox$).")$

bbox$\_$x $=$ bbox$\_$y $=0$

image$\_$width $=$ image$\_$height $=0$

category$\_$id $=-1$ # Assign a default category ID

# Check if the image is already added; if not, add it to the images list

if file$\_$name not in image$\_$id$\_$map:

image$\_$entry $=\{$

"file$\_$name": file$\_$name,

"height": image$\_$height,

"width": image$\_$width,

$"$id$"$: image$\_$id$\_$counter

$\}$

images.append$($image$\_$entry$)$

image$\_$id$\_$map$[$file$\_$name$]=$ image$\_$id$\_$counter

image$\_$id$\_$counter $+=1$

# Get the image$\_$id for this image

image$\_$id $=$ image$\_$id$\_$map$[$file$\_$name$]$

# Create entry for the bounding box annotation

annotation$\_$entry $=\{$

"image$\_$id$"$: image$\_$id$,$

"bbox": $[$bbox$\_$x$,$ bbox$\_$y$,$ bbox$\_$x $+$ image$\_$width, bbox$\_$y $+$ image$\_$height$],$

"category$\_$id$"$: category$\_$id

$\}$

annotations.append$($annotation$\_$entry$)$

# Combine images and annotations into the final dictionary

output$\_$dict $=\{$

"images": images,

"annotations": annotations

$\}$

# Write the output dictionary to the JSON file

with json$\_$file$\_$path.open$($mode$=\text{'}$w$\text{'})$ as f:

json.dump$($output$\_$dict, f$,$ indent$=4)$

print$($f$"$Successfully converted $\{$csv$\_$file$\_$path$\}$ to $\{$json$\_$file$\_$path$\}")$

# Specify your input and output file paths here

csv$\_$path $="$C:$\backslash \backslash$path$\backslash \backslash$to$\backslash \backslash$your$\backslash \backslash$annotations$.$csv$"$ # Change this to your CSV file path

json$\_$path $="$C:$\backslash \backslash$path$\backslash \backslash$to$\backslash \backslash$your$\backslash \backslash$output$.$json" # Change this to your desired JSON file path

# Run the conversion function

convert$\_$csv$\_$to$\_$json$($csv$\_$path, json$\_$path$)$