[FIXED] How to convert a CSV table into COCO format in python?

Issue

I have a CSV table with the following columns:

column_names = ['image_id', 'xmin', 'ymin', 'width', 'height', 'xmax','ymax']

where xmin, ymin, xmax and ymax represent the bounding box that encloses some object; width and height, the image dimensions; and image_id, the file name (.JPG file). Since I want to do object detection, I need to convert this table into COCO format. Amazingly enough, I can't find any answer to this inquiry in the internet.

Solution

I had the same issue before, then I found this code it is very helpful you will need to change the column names to this columns and update the csv file

column_names =['filename','class','width', 'height','xmin','ymin','xmax','ymax']

then try this code

import numpy as np
import json
import pandas as pd

path = 'annotations.csv' # the path to the CSV file
save_json_path = 'traincoco.json'


data = pd.read_csv(path)

images = []
categories = []
annotations = []

category = {}
category["supercategory"] = 'none'
category["id"] = 0
category["name"] = 'None'
categories.append(category)

data['fileid'] = data['filename'].astype('category').cat.codes
data['categoryid']= pd.Categorical(data['class'],ordered= True).codes
data['categoryid'] = data['categoryid']+1
data['annid'] = data.index

def image(row):
    image = {}
    image["height"] = row.height
    image["width"] = row.width
    image["id"] = row.fileid
    image["file_name"] = row.filename
    return image

def category(row):
    category = {}
    category["supercategory"] = 'None'
    category["id"] = row.categoryid
    category["name"] = row[2]
    return category

def annotation(row):
    annotation = {}
    area = (row.xmax -row.xmin)*(row.ymax - row.ymin)
    annotation["segmentation"] = []
    annotation["iscrowd"] = 0
    annotation["area"] = area
    annotation["image_id"] = row.fileid

    annotation["bbox"] = [row.xmin, row.ymin, row.xmax -row.xmin,row.ymax-row.ymin ]

    annotation["category_id"] = row.categoryid
    annotation["id"] = row.annid
    return annotation

for row in data.itertuples():
    annotations.append(annotation(row))

imagedf = data.drop_duplicates(subset=['fileid']).sort_values(by='fileid')
for row in imagedf.itertuples():
    images.append(image(row))

catdf = data.drop_duplicates(subset=['categoryid']).sort_values(by='categoryid')
for row in catdf.itertuples():
    categories.append(category(row))

data_coco = {}
data_coco["images"] = images
data_coco["categories"] = categories
data_coco["annotations"] = annotations
json.dump(data_coco, open(save_json_path, "w"), indent=4)

Answered By - Mohamed Fathy

Answer Checked By - David Marino (PHPFixing Volunteer)