2025-12-11 08:37:09 +08:00
|
|
|
|
|
|
|
|
|
|
# convert_yolo_to_cvat.py
|
|
|
|
|
|
import os
|
|
|
|
|
|
import xml.etree.ElementTree as ET
|
|
|
|
|
|
import cv2
|
|
|
|
|
|
|
|
|
|
|
|
# 输入 YOLO 标签目录
|
|
|
|
|
|
labels_dir = "keypoints_txt"
|
|
|
|
|
|
|
|
|
|
|
|
# 输出 XML 文件
|
|
|
|
|
|
output_xml = "annotations_cvat.xml"
|
|
|
|
|
|
|
|
|
|
|
|
# 图片目录（用于 width/height）
|
2026-03-10 13:58:21 +08:00
|
|
|
|
images_dir = "/media/hx/04e879fa-d697-4b02-ac7e-a4148876ebb0/dataset/1/20251230"
|
2025-12-11 08:37:09 +08:00
|
|
|
|
|
|
|
|
|
|
# 类别映射
|
|
|
|
|
|
class_mapping_reverse = {
|
|
|
|
|
|
0: 'clamp1',
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
def unnormalize_bbox(img_w, img_h, xc, yc, w, h):
|
|
|
|
|
|
"""YOLO 归一化 bbox 转回绝对坐标"""
|
|
|
|
|
|
cx = xc * img_w
|
|
|
|
|
|
cy = yc * img_h
|
|
|
|
|
|
bw = w * img_w
|
|
|
|
|
|
bh = h * img_h
|
|
|
|
|
|
|
|
|
|
|
|
x1 = cx - bw / 2
|
|
|
|
|
|
y1 = cy - bh / 2
|
|
|
|
|
|
x2 = cx + bw / 2
|
|
|
|
|
|
y2 = cy + bh / 2
|
|
|
|
|
|
return x1, y1, x2, y2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def unnormalize_points(img_w, img_h, kpts):
|
|
|
|
|
|
"""恢复关键点绝对坐标"""
|
|
|
|
|
|
pts = []
|
|
|
|
|
|
for i in range(0, len(kpts), 3):
|
|
|
|
|
|
x_n = float(kpts[i])
|
|
|
|
|
|
y_n = float(kpts[i + 1])
|
|
|
|
|
|
x = x_n * img_w
|
|
|
|
|
|
y = y_n * img_h
|
|
|
|
|
|
pts.append((x, y))
|
|
|
|
|
|
return pts
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# ----------------------- 构建 XML 树结构 -----------------------
|
|
|
|
|
|
|
|
|
|
|
|
root = ET.Element("annotations")
|
|
|
|
|
|
version = ET.SubElement(root, "version")
|
|
|
|
|
|
version.text = "1.1"
|
|
|
|
|
|
|
|
|
|
|
|
meta = ET.SubElement(root, "meta")
|
|
|
|
|
|
task = ET.SubElement(meta, "task")
|
|
|
|
|
|
ET.SubElement(task, "name").text = "converted_from_yolo"
|
|
|
|
|
|
# size 仍为总 txt 数量（即使部分无图）
|
|
|
|
|
|
txt_files = [f for f in os.listdir(labels_dir) if f.endswith(".txt")]
|
|
|
|
|
|
ET.SubElement(task, "size").text = str(len(txt_files))
|
|
|
|
|
|
|
|
|
|
|
|
# labels 元信息（增强兼容性）
|
|
|
|
|
|
labels_elem = ET.SubElement(task, "labels")
|
|
|
|
|
|
for name in class_mapping_reverse.values():
|
|
|
|
|
|
lab = ET.SubElement(labels_elem, "label")
|
|
|
|
|
|
ET.SubElement(lab, "name").text = name
|
|
|
|
|
|
ET.SubElement(lab, "color").text = "#ffffff"
|
|
|
|
|
|
ET.SubElement(lab, "type").text = "any"
|
|
|
|
|
|
ET.SubElement(lab, "attributes")
|
|
|
|
|
|
|
|
|
|
|
|
ET.SubElement(meta, "dumped").text = ""
|
|
|
|
|
|
|
|
|
|
|
|
# ----------------------- 处理每个 txt 文件（按顺序分配 id）-----------------------
|
|
|
|
|
|
sorted_txts = sorted(txt_files)
|
|
|
|
|
|
|
|
|
|
|
|
for idx, txt_file in enumerate(sorted_txts):
|
|
|
|
|
|
base = os.path.splitext(txt_file)[0]
|
|
|
|
|
|
img_path = os.path.join(images_dir, base + ".jpg")
|
|
|
|
|
|
|
|
|
|
|
|
# 尝试其他常见扩展名
|
|
|
|
|
|
img_found = False
|
|
|
|
|
|
for ext in [".jpg", ".jpeg", ".png", ".bmp"]:
|
|
|
|
|
|
candidate = os.path.join(images_dir, base + ext)
|
|
|
|
|
|
if os.path.exists(candidate):
|
|
|
|
|
|
img_path = candidate
|
|
|
|
|
|
img_found = True
|
|
|
|
|
|
break
|
|
|
|
|
|
|
|
|
|
|
|
if not img_found:
|
|
|
|
|
|
print(f"❗缺少图片：{base}.* （ID={idx} 被跳过）")
|
|
|
|
|
|
continue # 不生成 <image>，但 idx 已计入顺序
|
|
|
|
|
|
|
|
|
|
|
|
# 读取图片尺寸
|
|
|
|
|
|
img = cv2.imread(img_path)
|
|
|
|
|
|
if img is None:
|
|
|
|
|
|
print(f"❗无法读取图片：{img_path} （ID={idx} 被跳过）")
|
|
|
|
|
|
continue
|
|
|
|
|
|
|
|
|
|
|
|
h, w = img.shape[:2]
|
|
|
|
|
|
|
|
|
|
|
|
# 使用 enumerate 的 idx 作为 id（按 txt 顺序）
|
|
|
|
|
|
image_elem = ET.SubElement(
|
|
|
|
|
|
root, "image",
|
|
|
|
|
|
id=str(idx),
|
|
|
|
|
|
name=os.path.basename(img_path),
|
|
|
|
|
|
width=str(w),
|
|
|
|
|
|
height=str(h)
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# 读取 YOLO label
|
|
|
|
|
|
yolo_path = os.path.join(labels_dir, txt_file)
|
|
|
|
|
|
with open(yolo_path, "r") as f:
|
|
|
|
|
|
line = f.readline().strip()
|
|
|
|
|
|
if not line:
|
|
|
|
|
|
print(f"⚠ 空标签文件：{yolo_path}")
|
|
|
|
|
|
continue
|
|
|
|
|
|
parts = line.split()
|
|
|
|
|
|
|
|
|
|
|
|
cls_id = int(parts[0])
|
|
|
|
|
|
label_name = class_mapping_reverse[cls_id]
|
|
|
|
|
|
|
|
|
|
|
|
xc, yc, bw, bh = map(float, parts[1:5])
|
|
|
|
|
|
kpts = list(map(float, parts[5:]))
|
|
|
|
|
|
|
|
|
|
|
|
# 反归一化 bbox
|
|
|
|
|
|
x1, y1, x2, y2 = unnormalize_bbox(w, h, xc, yc, bw, bh)
|
|
|
|
|
|
|
|
|
|
|
|
# 反归一化关键点
|
|
|
|
|
|
pts = unnormalize_points(w, h, kpts)
|
|
|
|
|
|
|
|
|
|
|
|
# ---------------- CVAT <points> ----------------
|
|
|
|
|
|
pts_str = ";".join([f"{x:.2f},{y:.2f}" for x, y in pts])
|
|
|
|
|
|
ET.SubElement(
|
|
|
|
|
|
image_elem, "points",
|
|
|
|
|
|
label=label_name,
|
|
|
|
|
|
source="manual",
|
|
|
|
|
|
occluded="0",
|
|
|
|
|
|
points=pts_str,
|
|
|
|
|
|
z_order="0"
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
# ---------------- CVAT <box> ----------------
|
|
|
|
|
|
ET.SubElement(
|
|
|
|
|
|
image_elem, "box",
|
|
|
|
|
|
label=label_name,
|
|
|
|
|
|
source="manual",
|
|
|
|
|
|
occluded="0",
|
|
|
|
|
|
xtl=f"{x1:.2f}",
|
|
|
|
|
|
ytl=f"{y1:.2f}",
|
|
|
|
|
|
xbr=f"{x2:.2f}",
|
|
|
|
|
|
ybr=f"{y2:.2f}",
|
|
|
|
|
|
z_order="0"
|
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
# ----------------------- 写出 XML 文件 -----------------------
|
|
|
|
|
|
tree = ET.ElementTree(root)
|
|
|
|
|
|
tree.write(output_xml, encoding="utf-8", xml_declaration=True)
|
|
|
|
|
|
|
|
|
|
|
|
print("\n🎉 成功完成 YOLO → CVAT XML 转换！")
|
|
|
|
|
|
print(f"📁 输出文件：{output_xml}")
|
|
|
|
|
|
print(f"ℹ️ 共扫描 {len(sorted_txts)} 个标签文件，按其顺序分配 ID（缺失图片的 ID 已跳过）。")
|
