这篇文章提供了一个比较好的思路。总体做法还是离不开这个文章的做法。
上面文章作者提供的思路比较正统开始说明之前我们先说明一下coco2017的格式类型，这些网络上也说了很多了，这里再提及一下加深理解。
coco/
annotations/
instances_train2017.json
instances_val2017.json
images/
train2017/
0d4c5e4f-fc3c-4d5a-906c-105.jpg
val2017
0ddfc5aea-fcdac-421-92dad-144.jpg
以上格式摘自上面提到的文章，并且最后我们也是按这种方式生成的。
这里我们用到的格式也可以这么做，当然也可以适配自己的文件，比如要是每个视频下拆除的单张图，那么可以再多加一级目录出来。
具体的做法按照启发写到我的github项目里了，这里简要说明一下，当然github上后续也会更新。
csv格式制作
这里用csv也可以，用txt也可以，数据的格式并不是一个问题，而如何转换到我们想要的格式则有多种解决方式。
这里只展示csv的用法。
import os
import json
import cv2
import time
import argparse
import csv
parser = argparse.ArgumentParser(description='convert object label')
parser.add_argument('data', metavar='DIR',
help='path to dataset')
parser.add_argument('keyframe_dir', metavar='DIR',
help='path to frame dir')
parser.add_argument('--mode', type=str, choices=['train', 'val', 'test'])
args = parser.parse_args()
root = os.path.join(args.data, args.mode)
video_list = open(os.path.join(root, 'videolist.txt'),'r')
output_list = open('%s_list.txt' % args.mode,'w')
obj_name = open('objects_en.txt','r')
obj_list = [line.rstrip() for line in obj_name]
print(obj_list)
train_category = []
video_count = 0
start = time.time()
key_frame_count = 0
#output_folder = '%s_label' % args.mode
#if not os.path.exists(output_folder):
# os.makedirs(output_folder)
with open('val.csv','w') as f1:
for vid in video_list:
label = json.load(open(os.path.join(root, 'label', 'sample_' + vid.rstrip().split('.')[0] + '.json'), 'r'))
writer = csv.writer(f1)
for shot in label['shots']:
keyframe = shot['keyframe']
image_path = os.path.join(args.keyframe_dir, args.mode, vid[:-1], '%05d.jpg'% keyframe)
for target in shot['targets'] :
if target['category'] == 0:
if os.path.isfile(image_path):
xmin = target['bbox']['x']
xmax = xmin + target['bbox']['width']
ymin = target['bbox']['y']
ymax = ymin + target['bbox']['height']
cls_id = obj_list[int(target['tag'])]
writer.writerow([image_path,xmin,xmax,ymin,ymax,cls_id])#用write方法写入我们的csv文件里即可。
这里因为是适配了videonet比赛的数据格式，videonet数据集是一个视频数据集，目标检测也是其中一个任务，这里目标检测对应信息由官方给出的json来标注xyhw四个信息。我们csv的格式这里也是按
path xmin ymin xmax ymax id
上述格式生成csv文件，参考给出的代码也很好理解。通过csv文件将对应cocotrian或者validation格式的数据生成，经测试在maskrcnn-benchmark上是可以使用的。
以下给出train的coco形式的代码生成
# -*- coding: utf-8 -*-
'''
@time: 2019/01/11 11:28
spytensor
'''
import os
import json
import numpy as np
import pandas as pd
import glob
import cv2
import os
import shutil
from IPython import embed
from tqdm import tqdm
#from sklearn.model_selection import train_test_split
np.random.seed(41)
#0为背景
obj_name = open('objects_en.txt','r')
obj_list = [line.rstrip() for line in obj_name]
classname_to_id = {}
for i in range(len(obj_list)):
classname_to_id.setdefault(obj_list[i],i)
print(classname_to_id)
class Csv2CoCo:
def __init__(self,image_dir,total_annos):
self.images = []
self.annotations = []
self.categories = []
self.img_id = 0
self.ann_id = 0
self.image_dir = image_dir
self.total_annos = total_annos
def save_coco_json(self, instance, save_path):
json.dump(instance, open(save_path, 'w'), ensure_ascii=False, indent=2) # indent=2 更加美观显示
# 由txt文件构建COCO
def to_coco(self, keys):
self._init_categories()
for key in keys:
self.images.append(self._image(key))
shapes = self.total_annos[key]
for shape in shapes:
bboxi = []
for cor in shape[:-1]:
bboxi.append(int(cor))
label = shape[-1]
#print('label',label)
annotation = self._annotation(bboxi,label)
print('annotation',annotation)
self.annotations.append(annotation)
self.ann_id += 1
self.img_id += 1
instance = {}
instance['info'] = 'spytensor created'
instance['license'] = ['license']
instance['images'] = self.images
instance['annotations'] = self.annotations
instance['categories'] = self.categories
return instance
# 构建类别
def _init_categories(self):
for k, v in classname_to_id.items():
category = {}
category['id'] = v
category['name'] = k
self.categories.append(category)
# 构建COCO的image字段
def _image(self, path):
image = {}
print(path)
img = cv2.imread(self.image_dir + path)
image['height'] = img.shape[0]
image['width'] = img.shape[1]
image['id'] = self.img_id
image['file_name'] = path
return image
# 构建COCO的annotation字段
def _annotation(self, shape,label):
# label = shape[-1]
points = shape[:4]
annotation = {}
annotation['id'] = self.ann_id
annotation['image_id'] = self.img_id
annotation['category_id'] = int(classname_to_id[label])
#annotation['segmentation'] = self._get_seg(points)
annotation['bbox'] = self._get_box(points)
annotation['iscrowd'] = 0
annotation['area'] = 1.0
return annotation
# COCO的格式： [x1,y1,w,h] 对应COCO的bbox格式
def _get_box(self, points):
min_x = points[0]
min_y = points[1]
max_x = points[2]
max_y = points[3]
return [min_x, min_y, max_x - min_x, max_y - min_y]
# segmentation
def _get_seg(self, points):
min_x = points[0]
min_y = points[1]
max_x = points[2]
max_y = points[3]
h = max_y - min_y
w = max_x - min_x
a = []
a.append([min_x,min_y, min_x,min_y+0.5*h, min_x,max_y, min_x+0.5*w,max_y, max_x,max_y, max_x,max_y-0.5*h, max_x,min_y, max_x-0.5*w,min_y])
return a
if __name__ == '__main__':
csv_file = "train.csv"
image_dir = "/home/sda/videonet/train/image/train/"
saved_coco_path = "./"
# 整合csv格式标注文件
total_csv_annotations = {}
annotations = pd.read_csv(csv_file,header=None).values
for annotation in annotations:
#print(annotation[0].split(os.sep)[-2]+'/'+annotation[0].split(os.sep)[-1])
key = annotation[0].split(os.sep)[-2]+'/'+annotation[0].split(os.sep)[-1]
value = np.array([annotation[1:]])
if key in total_csv_annotations.keys():
total_csv_annotations[key] = np.concatenate((total_csv_annotations[key],value),axis=0)
else:
total_csv_annotations[key] = value
for k,v in total_csv_annotations.items():
print(k,v)
# 按照键值划分数据
total_keys = list(total_csv_annotations.keys())
train_keys = total_keys
#val_keys = total_keys
#train_keys, val_keys = train_test_split(total_keys, test_size=0.2)
print("train_n:", len(train_keys))#, 'val_n:', len(val_keys))
# 创建必须的文件夹
if not os.path.exists('%scoco/annotations/'%saved_coco_path):
os.makedirs('%scoco/annotations/'%saved_coco_path)
if not os.path.exists('%scoco/images/train2017/'%saved_coco_path):
os.makedirs('%scoco/images/train2017/'%saved_coco_path)
if not os.path.exists('%scoco/images/val2017/'%saved_coco_path):
os.makedirs('%scoco/images/val2017/'%saved_coco_path)
# 把训练集转化为COCO的json格式
for file in train_keys:
if not os.path.exists('{}coco/images/train2017/{}'.format(saved_coco_path,file.split('/')[0])):
print(file.split('/')[0])
os.makedirs('{}coco/images/train2017/{}'.format(saved_coco_path,file.split('/')[0]))
if not os.path.exists("{}coco/images/train2017/{}".format(saved_coco_path,file)):
shutil.copy(image_dir+file,"{}coco/images/train2017/{}".format(saved_coco_path,file))
elif os.path.exists('{}coco/images/train2017/{}'.format(saved_coco_path,file.split('/')[0])):
if not os.path.exists("{}coco/images/train2017/{}".format(saved_coco_path,file)):
shutil.copy(image_dir+file,"{}coco/images/train2017/{}".format(saved_coco_path,file))
l2c_train = Csv2CoCo(image_dir=image_dir,total_annos=total_csv_annotations)
train_instance = l2c_train.to_coco(train_keys)
l2c_train.save_coco_json(train_instance, '%scoco/annotations/instances_train2017.json'%saved_coco_path)
#for file in val_keys:
# shutil.copy(image_dir+file,"%scoco/images/val2017/"%saved_coco_path)
# 把验证集转化为COCO的json格式
#l2c_val = Csv2CoCo(image_dir=image_dir,total_annos=total_csv_annotations)
#val_instance = l2c_val.to_coco(val_keys)
#l2c_val.save_coco_json(val_instance, '%scoco/annotations/instances_val2017.json'%saved_coco_path)
和开头作者提到的方法不一样，我这里适配了自己的路径写到绝对路径，一定程度上保证了路径问题不会产生，新添加的多个if语句主要是有以下功能保证：
- 因为一张图片会包含多个目标物体，因此会可能有多个xyhw，当我们按上面生成csv代码时，因为一张图包含的多个xyhw，都会让所在的这张图重复写一遍，所以说我们用if先判断一遍看看是否是重复的，重复的我们跳过。
- 因为我们这里标注的是关键帧，所以说上一级目录是一个视频名，如果要把这个写到filename里，相当于我们还要再遍历一遍文件夹是否重复，所以再用到一次if判断。
对于train和validaiton是可以用一种方式生成的，而test会有一些变化，主要是test没有标注，我们只需要遍历后把filename写进去即可。