在创建gt database时报错

[Lucas-SJY]

在我按照readme中尝试使用kitti数据集时，使用python ./Paddle3D/tools/create_det_gt_database.py --dataset_name kitti --dataset_root ./kittimini --save_dir ./dataset/kitti命令创建gt database时报错

2024-01-11 16:29:09,128 - INFO - Begin to generate a database for the KITTI dataset.
libpng error: IDAT: bad parameters to zlib
Traceback (most recent call last):
File "/home/lucas/PycharmProjects/paddle3d/./Paddle3D/tools/create_det_gt_database.py", line 55, in
main(args)
File "/home/lucas/PycharmProjects/paddle3d/./Paddle3D/tools/create_det_gt_database.py", line 46, in main
generate_kitti_gt_database(args.dataset_root, save_dir=args.save_dir)
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle3d-1.0.0-py3.9.egg/paddle3d/datasets/generate_gt_database.py", line 52, in generate_kitti_gt_database
sample = dataset[data_idx]
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle3d-1.0.0-py3.9.egg/paddle3d/datasets/kitti/kitti_pointcloud_det.py", line 62, in getitem
sample = self.transforms(sample)
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle3d-1.0.0-py3.9.egg/paddle3d/transforms/base.py", line 45, in call
sample = t(sample)
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle3d-1.0.0-py3.9.egg/paddle3d/transforms/reader.py", line 185, in call
im_shape = cv2.imread(str(im_path)).shape[:2]
AttributeError: 'NoneType' object has no attribute 'shape'

可以确定路径没错，单独使用cv2.imread也能读取图片数据，在ai studio上也能顺利运行，但是在我的本地会发生如上报错

[warrentdrew]

您好，从报错看应该还是im_path不正确或是某些图片文件缺失导致imread没有读取到图片出现None的情况，建议在代码中增加一些im_path打印或者os.path.exists检查下路径和图片读取的正确性

[Lucas-SJY]

我的路径存在，且直接使用cv2.imread可以读取这些图片，print(cv2.imread("kittimini/training/image_2/000012.png").shape[:2])可以正常返回(375, 1242)
但是使用create_det_gt_database.py脚本依然返回None，且报错信息中有libpng error: IDAT: bad parameters to zlib
完整报错信息如下
-------------------current path kitti300frame/training/image_2/000012.png
Path exist
libpng error: IDAT: bad parameters to zlib
Traceback (most recent call last):
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/tools/create_det_gt_database.py", line 55, in
main(args)
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/tools/create_det_gt_database.py", line 46, in main
generate_kitti_gt_database(args.dataset_root, save_dir=args.save_dir)
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/paddle3d/datasets/generate_gt_database.py", line 52, in generate_kitti_gt_database
sample = dataset[data_idx]
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/paddle3d/datasets/kitti/kitti_pointcloud_det.py", line 62, in getitem
sample = self.transforms(sample)
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/paddle3d/transforms/base.py", line 45, in call
sample = t(sample)
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/paddle3d/transforms/reader.py", line 186, in call
im_shape = cv2.imread(str(im_path)).shape[:2]
AttributeError: 'NoneType' object has no attribute 'shape'

[warrentdrew]

可能跟环境有关，可以检查下opencv-python的版本以及依赖的libpng的版本，或者试下其他图像读取库如PIL Image.open()看下能否正常读取

[Lucas-SJY]

在我将opencv-python版本切换成4.5.5.64后，再次运行create_det_gt_database.py脚本，报错变成了
Traceback (most recent call last):
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/tools/create_det_gt_database.py", line 17, in
from paddle3d.datasets.generate_gt_database import (
File "/home/lucas/PycharmProjects/paddle3d/Paddle3D/paddle3d/init.py", line 17, in
import paddle
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle/init.py", line 75, in
from paddle import ( # noqa: F401
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle/dataset/init.py", line 18, in
from . import ( # noqa: F401
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle/dataset/flowers.py", line 36, in
from paddle.dataset.image import load_image_bytes, simple_transform
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/paddle/dataset/image.py", line 39, in
import cv2
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/cv2/init.py", line 190, in
bootstrap()
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/cv2/init.py", line 184, in bootstrap
if __load_extra_py_code_for_module("cv2", submodule, DEBUG):
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/cv2/init.py", line 37, in __load_extra_py_code_for_module
py_module = importlib.import_module(module_name)
File "/home/lucas/.conda/envs/paddle/lib/python3.9/importlib/init.py", line 127, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
File "/home/lucas/.conda/envs/paddle/lib/python3.9/site-packages/cv2/typing/init.py", line 156, in
Prim = _typing.Union[cv2.gapi.wip.draw.Text, cv2.gapi.wip.draw.Circle, cv2.gapi.wip.draw.Image, cv2.gapi.wip.draw.Line, cv2.gapi.wip.draw.Rect, cv2.gapi.wip.draw.Mosaic, cv2.gapi.wip.draw.Poly]
AttributeError: module 'cv2.gapi.wip.draw' has no attribute 'Text'

[warrentdrew]

这个应该和opencv的依赖有关，建议尝试下其他opencv的版本如4.6.0.66