mmdetection实战，训练自己的数据集
1 库安装
pip install timm==1.0.7 thop efficientnet_pytorch==0.7.1 einops grad-cam==1.4.8 dill==0.3.6 albumentations==1.4.11 pytorch_wavelets==1.3.0 tidecv PyWavelets -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install -U openmim -i https://pypi.tuna.tsinghua.edu.cn/simple
mim install mmengine -i https://pypi.tuna.tsinghua.edu.cn/simple
mim install "mmcv==2.1.0" -i https://pypi.tuna.tsinghua.edu.cn/simplepip install YOLO
pip install ultralytics*需要注意mmcv库的兼容性。
2 源码下载
GitHub - open-mmlab/mmdetection at v3.0.0
3 数据集准备
我是yolo数据集转coco
使用时需要修改类别和路径。
import json
import os
import shutil
import cv2
# info ，license，categories 结构初始化；
# 在train.json,val.json,test.json里面信息是一致的；
# info，license暂时用不到
info = {
"year": 2024,
"version": '1.0',
"date_created": 2024 - 6 - 9
}
licenses = {
"id": 1,
"name": "null",
"url": "null",
}
# 自己的标签类别，跟yolo的数据集类别要对应好；
categories = [
{
"id": 0,
"name": 'L',
"supercategory": 'lines',
},
{
"id": 1,
"name": 'R',
"supercategory": 'lines',
},
{
"id": 2,
"name": 'I',
"supercategory": 'lines',
},
{
"id": 3,
"name": 'M',
"supercategory": 'lines',
},
{
"id": 4,
"name": 'A',
"supercategory": 'lines',
}
]
# 初始化train,test、valid 数据字典
# info licenses categories 在train和test里面都是一致的；
train_data = {'info': info, 'licenses': licenses, 'categories': categories, 'images': [], 'annotations': []}
test_data = {'info': info, 'licenses': licenses, 'categories': categories, 'images': [], 'annotations': []}
valid_data = {'info': info, 'licenses': licenses, 'categories': categories, 'images': [], 'annotations': []}
# image_path 对应yolov8的图像路径，比如images/train；
# label_path 对应yolov8的label路径，比如labels/train 跟images要对应；
def yolo_covert_coco_format(image_path, label_path):
images = []
annotations = []
for index, img_file in enumerate(os.listdir(image_path)):
if img_file.endswith('.jpg'):
image_info = {}
img = cv2.imread(os.path.join(image_path, img_file))
height, width, channel = img.shape
image_info['id'] = index
image_info['file_name'] = img_file
image_info['width'], image_info['height'] = width, height
else:
continue
if image_info != {}:
images.append(image_info)
# 处理label信息-------
label_file = os.path.join(label_path, img_file.replace('.jpg', '.txt'))
with open(label_file, 'r') as f:
for idx, line in enumerate(f.readlines()):
info_annotation = {}
class_num, xs, ys, ws, hs = line.strip().split(' ')
class_id, xc, yc, w, h = int(class_num), float(xs), float(ys), float(ws), float(hs)
xmin = (xc - w / 2) * width
ymin = (yc - h / 2) * height
xmax = (xc + w / 2) * width
ymax = (yc + h / 2) * height
bbox_w = int(width * w)
bbox_h = int(height * h)
img_copy = img[int(ymin):int(ymax), int(xmin):int(xmax)].copy()
info_annotation["category_id"] = class_id # 类别的id
info_annotation['bbox'] = [xmin, ymin, bbox_w, bbox_h] ## bbox的坐标
info_annotation['area'] = bbox_h * bbox_w ###area
info_annotation['image_id'] = index # bbox的id
info_annotation['id'] = index * 100 + idx # bbox的id
# cv2.imwrite(f"./temp/{info_annotation['id']}.jpg", img_copy)
info_annotation['segmentation'] = [[xmin, ymin, xmax, ymin, xmax, ymax, xmin, ymax]] # 四个点的坐标
info_annotation['iscrowd'] = 0 # 单例
annotations.append(info_annotation)
return images, annotations
# key == train，test，val
# 对应要生成的json文件，比如instances_train.json，instances_test.json，instances_val.json
# 只是为了不重复写代码。。。。。
def gen_json_file(yolov8_data_path, coco_format_path, key):
# json path
json_path = os.path.join(coco_format_path, f'annotations/instances_{key}.json')
dst_path = os.path.join(coco_format_path, f'{key}')
if not os.path.exists(os.path.dirname(json_path)):
os.makedirs(os.path.dirname(json_path), exist_ok=True)
data_path = os.path.join(yolov8_data_path, f'images/{key}')
label_path = os.path.join(yolov8_data_path, f'labels/{key}')
images, anns = yolo_covert_coco_format(data_path, label_path)
if key == 'train':
train_data['images'] = images
train_data['annotations'] = anns
with open(json_path, 'w') as f:
json.dump(train_data, f, indent=2)
# shutil.copy(data_path,'')
elif key == 'test':
test_data['images'] = images
test_data['annotations'] = anns
with open(json_path, 'w') as f:
json.dump(test_data, f, indent=2)
elif key == 'val':
valid_data['images'] = images
valid_data['annotations'] = anns
with open(json_path, 'w') as f:
json.dump(valid_data, f, indent=2)
else:
print(f'key is {key}')
print(f'generate {key} json success!')
return
if __name__ == '__main__':
yolov8_data_path = r''
coco_format_path = r''
gen_json_file(yolov8_data_path, coco_format_path, key='train')
gen_json_file(yolov8_data_path, coco_format_path, key='val')
gen_json_file(yolov8_data_path, coco_format_path, key='test')coco标签可视化工具，可以用于检查数据集是否有问题
import matplotlib.pyplot as plt
import matplotlib.patches as patches
from pycocotools.coco import COCO
import numpy as np
import skimage.io as io
import os
# 定义COCO数据集的路径
dataDir = r''
dataType = 'train2017'
annFile = f'{dataDir}/annotations/instances_{dataType}.json'
# 初始化COCO API
coco = COCO(annFile)
# 获取一张图片的ID
imgIds = coco.getImgIds()
img = coco.loadImgs(imgIds[np.random.randint(0, len(imgIds))])[0]
# 加载并显示图片
I = io.imread(f'{dataDir}/{dataType}/{img["file_name"]}')
plt.imshow(I)
plt.axis('off')
# 获取图片中的标注
annIds = coco.getAnnIds(imgIds=img['id'], iscrowd=None)
anns = coco.loadAnns(annIds)
# 显示标注和类别标签
for ann in anns:
bbox = ann['bbox']
category_id = ann['category_id']
category_name = coco.loadCats(category_id)[0]['name'] # 获取类别名称
# 画出边框
rect = patches.Rectangle((bbox[0], bbox[1]), bbox[2], bbox[3], linewidth=2, edgecolor='r', facecolor='none')
plt.gca().add_patch(rect)
# 在边框上方显示类别名称
plt.text(bbox[0], bbox[1] - 10, category_name, color='yellow', fontsize=12, weight='bold', backgroundcolor='black')
# 确保输出目录存在
output_path = f'./{img["file_name"]}_annotated.jpg'
plt.savefig(output_path, bbox_inches='tight', pad_inches=0.1)
plt.show()
4 代码修改
命令行进入代码文件夹，然后输入指令
pip install -v -e.根据你想要跑什么模型，以faster rcnn为例。
你需要修改faster-rcnn_r50_fpn_1x.py ,classes_name.py ,coco.py, coco_detection.py 和shedule_1x.py中的数据集参数和训练参数。例如类别名，类别数，数据集路径，epoches和batch等。
这些文件的路径
mmdetection-3.0.0/configs/_base_/schedules/schedule_1x.pymmdetection-3.0.0/configs/_base_/datasets/coco_detection.pymmdetection-3.0.0/configs/_base_/models/faster-rcnn_r50_fpn.pymmdetection-3.0.0/mmdet/datasets/coco.py5 运行
python tools/train.py configs/_base_/models/faster-rcnn_r50_fpn.py或
python tools/train.py work_dirs/faster-rcnn_r50_fpn_1x_coco/faster-rcnn_r50_fpn_1x_coco.py后续若需要修改参数可以直接在faster-rcnn_r50_fpn_1x_coco.py中修改。
6 测试
python tools/test.py work_dirs/faster-rcnn_r50_fpn_1x_coco/faster-rcnn_r50_fpn_1x_coco.py work_dirs/faster-rcnn_r50_fpn_1x_coco/epoch_20.pth --out res.pkl7 评价指标转换
来自
https://github.com/z1069614715
的代码。可以转换mmdet评价指标为yolo评价指标。
import os, torch, cv2, math, tqdm, time, shutil, argparse, json, pickle
import numpy as np
from prettytable import PrettyTable
def clip_boxes(boxes, shape):
# Clip boxes (xyxy) to image shape (height, width)
if isinstance(boxes, torch.Tensor): # faster individually
boxes[..., 0].clamp_(0, shape[1]) # x1
boxes[..., 1].clamp_(0, shape[0]) # y1
boxes[..., 2].clamp_(0, shape[1]) # x2
boxes[..., 3].clamp_(0, shape[0]) # y2
else: # np.array (faster grouped)
boxes[..., [0, 2]] = boxes[..., [0, 2]].clip(0, shape[1]) # x1, x2
boxes[..., [1, 3]] = boxes[..., [1, 3]].clip(0, shape[0]) # y1, y2
def scale_boxes(img1_shape, boxes, img0_shape, ratio_pad=None):
# Rescale boxes (xyxy) from img1_shape to img0_shape
if ratio_pad is None: # calculate from img0_shape
gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1]) # gain = old / new
pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2 # wh padding
else:
gain = ratio_pad[0][0]
pad = ratio_pad[1]
boxes[..., [0, 2]] -= pad[0] # x padding
boxes[..., [1, 3]] -= pad[1] # y padding
boxes[..., :4] /= gain
clip_boxes(boxes, img0_shape)
return boxes
def box_iou(box1, box2, eps=1e-7):
"""
Calculate intersection-over-union (IoU) of boxes. Both sets of boxes are expected to be in (x1, y1, x2, y2) format.
Based on https://github.com/pytorch/vision/blob/master/torchvision/ops/boxes.py
Args:
box1 (torch.Tensor): A tensor of shape (N, 4) representing N bounding boxes.
box2 (torch.Tensor): A tensor of shape (M, 4) representing M bounding boxes.
eps (float, optional): A small value to avoid division by zero. Defaults to 1e-7.
Returns:
(torch.Tensor): An NxM tensor containing the pairwise IoU values for every element in box1 and box2.
"""
# NOTE: Need .float() to get accurate iou values
# inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2)
(a1, a2), (b1, b2) = box1.float().unsqueeze(1).chunk(2, 2), box2.float().unsqueeze(0).chunk(2, 2)
inter = (torch.min(a2, b2) - torch.max(a1, b1)).clamp_(0).prod(2)
# IoU = inter / (area1 + area2 - inter)
return inter / ((a2 - a1).prod(2) + (b2 - b1).prod(2) - inter + eps)
def process_batch(detections, labels, iouv):
"""
Return correct prediction matrix
Arguments:
detections (array[N, 6]), x1, y1, x2, y2, conf, class
labels (array[M, 5]), class, x1, y1, x2, y2
Returns:
correct (array[N, 10]), for 10 IoU levels
"""
correct = np.zeros((detections.shape[0], iouv.shape[0])).astype(bool)
iou = box_iou(labels[:, 1:], detections[:, :4])
correct_class = labels[:, 0:1] == detections[:, 5]
for i in range(len(iouv)):
x = torch.where((iou >= iouv[i]) & correct_class) # IoU > threshold and classes match
if x[0].shape[0]:
matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy() # [label, detect, iou]
if x[0].shape[0] > 1:
matches = matches[matches[:, 2].argsort()[::-1]]
matches = matches[np.unique(matches[:, 1], return_index=True)[1]]
# matches = matches[matches[:, 2].argsort()[::-1]]
matches = matches[np.unique(matches[:, 0], return_index=True)[1]]
correct[matches[:, 1].astype(int), i] = True
return torch.tensor(correct, dtype=torch.bool, device=iouv.device)
def smooth(y, f=0.05):
# Box filter of fraction f
nf = round(len(y) * f * 2) // 2 + 1 # number of filter elements (must be odd)
p = np.ones(nf // 2) # ones padding
yp = np.concatenate((p * y[0], y, p * y[-1]), 0) # y padded
return np.convolve(yp, np.ones(nf) / nf, mode='valid') # y-smoothed
def ap_per_class(tp, conf, pred_cls, target_cls, plot=False, save_dir='.', names=(), eps=1e-16, prefix=''):
""" Compute the average precision, given the recall and precision curves.
Source: https://github.com/rafaelpadilla/Object-Detection-Metrics.
# Arguments
tp: True positives (nparray, nx1 or nx10).
conf: Objectness value from 0-1 (nparray).
pred_cls: Predicted object classes (nparray).
target_cls: True object classes (nparray).
plot: Plot precision-recall curve at mAP@0.5
save_dir: Plot save directory
# Returns
The average precision as computed in py-faster-rcnn.
"""
# Sort by objectness
i = np.argsort(-conf)
tp, conf, pred_cls = tp[i], conf[i], pred_cls[i]
# Find unique classes
unique_classes, nt = np.unique(target_cls, return_counts=True)
nc = unique_classes.shape[0] # number of classes, number of detections
# Create Precision-Recall curve and compute AP for each class
px, py = np.linspace(0, 1, 1000), [] # for plotting
ap, p, r = np.zeros((nc, tp.shape[1])), np.zeros((nc, 1000)), np.zeros((nc, 1000))
for ci, c in enumerate(unique_classes):
i = pred_cls == c
n_l = nt[ci] # number of labels
n_p = i.sum() # number of predictions
if n_p == 0 or n_l == 0:
continue
# Accumulate FPs and TPs
fpc = (1 - tp[i]).cumsum(0)
tpc = tp[i].cumsum(0)
# Recall
recall = tpc / (n_l + eps) # recall curve
r[ci] = np.interp(-px, -conf[i], recall[:, 0], left=0) # negative x, xp because xp decreases
# Precision
precision = tpc / (tpc + fpc) # precision curve
p[ci] = np.interp(-px, -conf[i], precision[:, 0], left=1) # p at pr_score
# AP from recall-precision curve
for j in range(tp.shape[1]):
ap[ci, j], mpre, mrec = compute_ap(recall[:, j], precision[:, j])
if plot and j == 0:
py.append(np.interp(px, mrec, mpre)) # precision at mAP@0.5
# Compute F1 (harmonic mean of precision and recall)
f1 = 2 * p * r / (p + r + eps)
i = smooth(f1.mean(0), 0.1).argmax() # max F1 index
p, r, f1 = p[:, i], r[:, i], f1[:, i]
tp = (r * nt).round() # true positives
fp = (tp / (p + eps) - tp).round() # false positives
return tp, fp, p, r, f1, ap, unique_classes.astype(int)
def compute_ap(recall, precision):
""" Compute the average precision, given the recall and precision curves
# Arguments
recall: The recall curve (list)
precision: The precision curve (list)
# Returns
Average precision, precision curve, recall curve
"""
# Append sentinel values to beginning and end
mrec = np.concatenate(([0.0], recall, [1.0]))
mpre = np.concatenate(([1.0], precision, [0.0]))
# Compute the precision envelope
mpre = np.flip(np.maximum.accumulate(np.flip(mpre)))
# Integrate area under curve
method = 'interp' # methods: 'continuous', 'interp'
if method == 'interp':
x = np.linspace(0, 1, 101) # 101-point interp (COCO)
ap = np.trapz(np.interp(x, mrec, mpre), x) # integrate
else: # 'continuous'
i = np.where(mrec[1:] != mrec[:-1])[0] # points where x axis (recall) changes
ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1]) # area under curve
return ap, mpre, mrec
def parse_opt():
parser = argparse.ArgumentParser()
parser.add_argument('--label_coco', type=str, default='/home/hjj/Desktop/dataset/dataset_visdrone/test_coco.json',
help='label coco path')
parser.add_argument('--pred_coco', type=str, default='runs/val/exp/predictions.json', help='pred coco path')
# parser.add_argument('--pred_coco', type=str, default='/home/hjj/Desktop/github_code/mmdetection-visdrone/work_dirs/dino-4scale_r50_8xb2-12e_visdrone/test/prediction.pickle', help='pred coco path')
parser.add_argument('--iou', type=float, default=0.7, help='iou threshold')
parser.add_argument('--conf', type=float, default=0.001, help='conf threshold')
opt = parser.parse_known_args()[0]
return opt
if __name__ == '__main__':
opt = parse_opt()
iouv = torch.linspace(0.5, 0.95, 10) # iou vector for mAP@0.5:0.95
niou = iouv.numel()
stats = []
label_coco_json_path, pred_coco_json_path = opt.label_coco, opt.pred_coco
with open(label_coco_json_path) as f:
label = json.load(f)
classes = []
for data in label['categories']:
classes.append(data['name'])
image_id_hw_dict = {}
for data in label['images']:
image_id_hw_dict[data['id']] = [data['height'], data['width']]
label_id_dict = {}
for data in tqdm.tqdm(label['annotations'], desc='Process label...'):
if data['image_id'] not in label_id_dict:
label_id_dict[data['image_id']] = []
category_id = data['category_id']
x_min, y_min, w, h = data['bbox'][0], data['bbox'][1], data['bbox'][2], data['bbox'][3]
x_max, y_max = x_min + w, y_min + h
label_id_dict[data['image_id']].append(np.array([int(category_id), x_min, y_min, x_max, y_max]))
if pred_coco_json_path.endswith('json'):
with open(pred_coco_json_path) as f:
pred = json.load(f)
pred_id_dict = {}
for data in tqdm.tqdm(pred, desc='Process pred...'):
if data['image_id'] not in pred_id_dict:
pred_id_dict[data['image_id']] = []
score = data['score']
category_id = data['category_id']
x_min, y_min, w, h = data['bbox'][0], data['bbox'][1], data['bbox'][2], data['bbox'][3]
x_max, y_max = x_min + w, y_min + h
pred_id_dict[data['image_id']].append(
np.array([x_min, y_min, x_max, y_max, float(score), int(category_id)]))
else:
with open(pred_coco_json_path, 'rb') as f:
pred = pickle.load(f)
pred_id_dict = {}
for data in tqdm.tqdm(pred, desc='Process pred...'):
image_id = os.path.splitext(os.path.basename(data['img_path']))[0]
if image_id not in pred_id_dict:
pred_id_dict[image_id] = []
for i in range(data['pred_instances']['labels'].size(0)):
score = data['pred_instances']['scores'][i]
category_id = data['pred_instances']['labels'][i]
bboxes = data['pred_instances']['bboxes'][i]
x_min, y_min, x_max, y_max = bboxes.cpu().detach().numpy()
# x_min, x_max = x_min / data['scale_factor'][0], x_max / data['scale_factor'][0]
# y_min, y_max = y_min / data['scale_factor'][1], y_max / data['scale_factor'][1]
pred_id_dict[image_id].append(np.array([x_min, y_min, x_max, y_max, float(score), int(category_id)]))
for idx, image_id in enumerate(tqdm.tqdm(list(image_id_hw_dict.keys()), desc="Cal mAP...")):
label = np.array(label_id_dict[image_id])
if image_id not in pred_id_dict:
pred = np.empty((0, 6))
else:
pred = torch.from_numpy(np.array(pred_id_dict[image_id]))
nl, npr = label.shape[0], pred.shape[0]
correct = torch.zeros(npr, niou, dtype=torch.bool)
if npr == 0:
if nl:
stats.append((correct, *torch.zeros((2, 0)), torch.from_numpy(label[:, 0])))
continue
if nl:
correct = process_batch(pred, torch.from_numpy(label), iouv)
stats.append((correct, pred[:, 4], pred[:, 5], torch.from_numpy(label[:, 0])))
stats = [torch.cat(x, 0).cpu().numpy() for x in zip(*stats)]
tp, fp, p, r, f1, ap, ap_class = ap_per_class(*stats)
print(f'precision:{p}')
print(f'recall:{r}')
print(f'mAP@0.5:{ap[:, 0]}')
table = PrettyTable()
table.title = f"Metrice"
table.field_names = ["Classes", 'Precision', 'Recall', 'mAP50', 'mAP50-95']
table.add_row(['all', f'{np.mean(p):.3f}', f'{np.mean(r):.3f}', f'{np.mean(ap[:, 0]):.3f}', f'{np.mean(ap):.3f}'])
for cls_idx, classes in enumerate(classes):
table.add_row([classes, f'{p[cls_idx]:.3f}', f'{r[cls_idx]:.3f}', f'{ap[cls_idx, 0]:.3f}',
f'{ap[cls_idx, :].mean():.3f}'])
print(table)需要使用coco格式的标签文件和推理生成的pkl文件。
转换后结果示例：
precision:[0.79133871 0.92334536 0.94697749 0.95896188 0.93151955]
recall:[0.7721585 0.9375 0.96192053 0.81843044 0.91293532]
mAP@0.5:[0.81393383 0.96300486 0.98147308 0.91452196 0.95637408]
+-------------------------------------------------+
| Metrice |
+---------+-----------+--------+-------+----------+
| Classes | Precision | Recall | mAP50 | mAP50-95 |
+---------+-----------+--------+-------+----------+
| all | 0.910 | 0.881 | 0.926 | 0.553 |
| LGB | 0.791 | 0.772 | 0.814 | 0.462 |
| RFB | 0.923 | 0.938 | 0.963 | 0.545 |
| IMM | 0.947 | 0.962 | 0.981 | 0.592 |
| MW | 0.959 | 0.818 | 0.915 | 0.567 |
| AGM | 0.932 | 0.913 | 0.956 | 0.596 |
+---------+-----------+--------+-------+----------+若出现评价指标为0的情况，建议检查标签文件是否有问题。
可以尝试重新转换一下标签文件。
yolo2coco
import os
import cv2
import json
from tqdm import tqdm
from sklearn.model_selection import train_test_split
import argparse
# visdrone2019
classes = ['LGB',
'RFB',
'IMM',
'MW',
'AGM']
parser = argparse.ArgumentParser()
parser.add_argument('--image_path', default=r'',type=str, help="path of images")
parser.add_argument('--label_path', default=r'',type=str, help="path of labels .txt")
parser.add_argument('--save_path', default='test.json', type=str, help="if not split the dataset, give a path to a json file")
arg = parser.parse_args()
def yolo2coco(arg):
print("Loading data from ", arg.image_path, arg.label_path)
assert os.path.exists(arg.image_path)
assert os.path.exists(arg.label_path)
originImagesDir = arg.image_path
originLabelsDir = arg.label_path
# images dir name
indexes = os.listdir(originImagesDir)
dataset = {'categories': [], 'annotations': [], 'images': []}
for i, cls in enumerate(classes, 0):
dataset['categories'].append({'id': i, 'name': cls, 'supercategory': 'mark'})
# 标注的id
ann_id_cnt = 0
for k, index in enumerate(tqdm(indexes)):
# 支持 png jpg 格式的图片.
txtFile = f'{index[:index.rfind(".")]}.txt'
stem = index[:index.rfind(".")]
# 读取图像的宽和高
try:
im = cv2.imread(os.path.join(originImagesDir, index))
height, width, _ = im.shape
except Exception as e:
print(f'{os.path.join(originImagesDir, index)} read error.\nerror:{e}')
# 添加图像的信息
if not os.path.exists(os.path.join(originLabelsDir, txtFile)):
# 如没标签，跳过，只保留图片信息.
continue
dataset['images'].append({'file_name': index,
'id': stem,
'width': width,
'height': height})
with open(os.path.join(originLabelsDir, txtFile), 'r') as fr:
labelList = fr.readlines()
for label in labelList:
label = label.strip().split()
x = float(label[1])
y = float(label[2])
w = float(label[3])
h = float(label[4])
# convert x,y,w,h to x1,y1,x2,y2
H, W, _ = im.shape
x1 = (x - w / 2) * W
y1 = (y - h / 2) * H
x2 = (x + w / 2) * W
y2 = (y + h / 2) * H
# 标签序号从0开始计算, coco2017数据集标号混乱，不管它了。
cls_id = int(label[0])
width = max(0, x2 - x1)
height = max(0, y2 - y1)
dataset['annotations'].append({
'area': width * height,
'bbox': [x1, y1, width, height],
'category_id': cls_id,
'id': ann_id_cnt,
'image_id': stem,
'iscrowd': 0,
# mask, 矩形是从左上角点按顺时针的四个顶点
'segmentation': [[x1, y1, x2, y1, x2, y2, x1, y2]]
})
ann_id_cnt += 1
# 保存结果
with open(arg.save_path, 'w') as f:
json.dump(dataset, f)
print('Save annotation to {}'.format(arg.save_path))
if __name__ == "__main__":
yolo2coco(arg)