def __init__(self, dataset_dir=None):
if dataset_dir:
self.paths = ConfigPaths(dataset_dir)
else:
self.paths = ConfigPaths(dtst_dir)
def __init__(self):
self.paths = ConfigPaths(dataset_dir)
import pandas as pd
from os.path import join
from argparse import ArgumentParser
"""
YOLO has 3 types of file:
1. image.jpg
2. image.txt --> 5 columns ==> class, xmin, ymin, w, h | normalized
3. classes.txt
"""
if __name__ == '__main__':
parser = ArgumentParser()
parser.add_argument('--dataset_dir', help='Root folder of dataset.')
dataset_dir = ''
paths = ConfigPaths(dataset_dir)
df = pd.read_csv(join(paths.standard_format_dir, 'dataset.csv'),
index_col=0)
for index, row in df.iterrows():
frame_name = row['frame_name']
classes = eval(row['class'])
xmins = eval(row['xmin'])
ymins = eval(row['ymin'])
xmaxs = eval(row['xmax'])
ymaxs = eval(row['ymax'])
width = row['width']
height = row['height']
with open(
join(paths.yolo_v5_pytorch_labels_dir,
from object_detection_dataset.constants._path_creator import ConfigPaths
if __name__ == '__main__':
root_folder = 'E:\\Data Sets\\Detection\\Maritime\\COCO x SEAGULL\\'
paths = ConfigPaths(root_folder, make_dirs=True)
categories, supercategories = [], []
for category_id in category_ids:
for val in self.data['categories']:
if val['id'] == category_id:
categories.append(val['name'])
supercategories.append(val['supercategory'])
return categories, supercategories
def __find_bbox_of_id(self, image_id: int):
total_bboxes = []
category_ids = []
id_list = []
for val in self.data['annotations']:
if val['image_id'] == image_id:
total_bboxes.append(val['bbox'])
category_ids.append(val['category_id'])
id_list.append(val['id'])
return total_bboxes, category_ids, id_list
if __name__ == '__main__':
dataset_dir = 'E:\Data Sets\Detection\COCO\outputs\original_format'
# dataset_dir = 'D:\Documents\Computer Vision\Object Detection\Datasets\COCO'
paths = ConfigPaths(dataset_dir, False)
json_path = os_path_join(dataset_dir, 'instances_train2017.json')
handler = COCO_JSON_Handler(dataset_dir, json_path, extract_data=True)
handler.set_json_dir(dataset_dir)
handler.show_all_categories()
handler.extract_categories(dataset_dir)
import os
import pandas as pd
from object_detection_dataset.constants._path_creator import ConfigPaths
import shutil
if __name__ == '__main__':
coco__folder = 'E:\\Data Sets\\Detection\\COCO\\'
coco_paths = ConfigPaths(coco__folder, make_dirs=True)
coco_x_seagull_folder = 'E:\\Data Sets\\Detection\\Maritime\\COCO x SEAGULL\\'
coco_x_seagull_paths = ConfigPaths(coco_x_seagull_folder, make_dirs=True)
coco_x_seagull_df = pd.read_csv(os.path.join(coco_x_seagull_paths.standard_format_dir, 'coco_x_seagull.csv'))
for index, row in coco_x_seagull_df.iterrows():
image_name = row['filename']
src = os.path.join(coco_paths.images_dir, 'train2017', image_name)
dst = os.path.join(coco_x_seagull_paths.images_dir, image_name)
print('copying', image_name)
print('From:', src)
print('To:', dst)
shutil.copy(src, dst)
def __init__(self, dataset_dir: str):
self.paths = ConfigPaths(dataset_dir, make_dirs=False)
self.standard_csv_path = os_path_join(self.paths.standard_format_dir, 'all.csv')