sidiangongyuan commented 5 months ago

In nuscenes_dataset.py

NOTE load image annotations

    gt_bboxes_2d = info['gt_boxes_2d'][mask]
    gt_labels_2d = gt_labels_3d.copy()

KeyError: 'gt_boxes_2d'.

what's wrong ?

yinjunbo commented 5 months ago

Hi, please follow the code in AutoAlignV2 exactly for generting the train/val infos.

sidiangongyuan commented 4 months ago

Here is my data structure, is there any problem?

I just run python tools/create_data.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag nuscenes

The extra content is generated by other tasks and should not affect .

yinjunbo commented 4 months ago

Hi, thanks for your interest. If you've re-generated the infos from AutoAlignV2, 'gt_boxes_2d' should have been contained in the infos, according to _https://github.com/yinjunbo/IS-Fusion/blob/ef65e0a0e21eba9a86e1ca209492e35bbfbe7966/tools/data_converter/nuscenes_converter.py#L350_

sidiangongyuan commented 4 months ago

python tools/create_data.py nuscenes --root-path ./data/nuscenes --out-dir ./data/nuscenes --extra-tag nuscenes will not generate train and val .pkl , it just generate test.pkl.

and do I need to uncomment this?

yinjunbo commented 4 months ago

Hi, you can uncomment this https://github.com/yinjunbo/IS-Fusion/blob/b79667a4ca5e857fe3909864d87606c22c3804a7/tools/create_data.py#L230 to geenrate train&val infos.

sidiangongyuan commented 4 months ago

Yes I found that.

And do I need to uncomment codes I show before ?

yinjunbo commented 4 months ago

No need. But please remember to back up your previous infos to avoid it being overwritten by the new ones.

sidiangongyuan commented 4 months ago

OK， thank you for your help .

sidiangongyuan commented 4 months ago

No need. But please remember to back up your previous infos to avoid it being overwritten by the new ones.

Sorry, I encountered a problem again, when I generated the pkl of train and val,

FileNotFoundError: NuScenesDataset: [Errno 2] No such file or directory: './data/nuscenes/nuscenes_infos_trainval.pkl'

def nuscenes_data_prep(root_path, info_prefix, version, dataset_name, out_dir, max_sweeps=10): """Prepare data related to nuScenes dataset.

Related data consists of '.pkl' files recording basic infos,
2D annotations and groundtruth database.

Args:
    root_path (str): Path of dataset root.
    info_prefix (str): The prefix of info filenames.
    version (str): Dataset version.
    dataset_name (str): The dataset class name.
    out_dir (str): Output directory of the groundtruth database info.
    max_sweeps (int): Number of input consecutive frames. Default: 10
"""
nuscenes_converter.create_nuscenes_infos(
    root_path, info_prefix, version=version, max_sweeps=max_sweeps)

if version == 'v1.0-test':
    info_test_path = osp.join(root_path, f'{info_prefix}_infos_test.pkl')
    nuscenes_converter.export_2d_annotation(
        root_path, info_test_path, version=version)
    return

info_train_path = osp.join(root_path, f'{info_prefix}_infos_train.pkl')
info_val_path = osp.join(root_path, f'{info_prefix}_infos_val.pkl')

# nuscenes_converter.export_2d_annotation(
#     root_path, info_train_path, version=version)
# nuscenes_converter.export_2d_annotation(
#     root_path, info_val_path, version=version)

create_groundtruth_database(dataset_name, root_path, info_prefix,
                            f'{out_dir}/{info_prefix}_infos_trainval.pkl',
                            with_bbox=True)

I need to confirm again, is the initial structure of the file like mine?

yinjunbo commented 4 months ago

Yes, the original files are like what you've showed. You can use the updated code to avoid the problem: https://github.com/yinjunbo/IS-Fusion/blob/167dc78de50da583a87597d57791f3dcbca91e9b/tools/create_data.py#L83C33-L83C76

sidiangongyuan commented 4 months ago

Thank you for your patient answer, I have solved the problem

loserhou commented 3 months ago

Thank you for your patient answer, I have solved the problem 请问现在是直接使用作者这个命令就可以了吗，还需要去[AutoAlignV2]生成吗？

sidiangongyuan commented 3 months ago

Thank you for your patient answer, I have solved the problem 请问现在是直接使用作者这个命令就可以了吗，还需要去[AutoAlignV2]生成吗？

是的

loserhou commented 3 months ago

谢谢！

	侯业飞 @.***

---- Replied Message ---- | From | @.> | | Date | 7/24/2024 10:02 | | To | @.> | | Cc | @.>, @.> | | Subject | Re: [yinjunbo/IS-Fusion] gt_bboxes_2d = info['gt_boxes_2d'][mask] (Issue #11) |

Thank you for your patient answer, I have solved the problem 请问现在是直接使用作者这个命令就可以了吗，还需要去[AutoAlignV2]生成吗？

是的

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

xqk666 commented 2 months ago

Hello, I encountered the same error when using the nuscenes mini dataset. I tried the above method but it didn't work. How can I fix it?

sidiangongyuan commented 1 month ago

Hello, I encountered the same error when using the nuscenes mini dataset. I tried the above method but it didn't work. How can I fix it?

我这里并没有使用 mini。使用的是完整的数据集。在生成数据的时候， train_version = f'{args.version}-trainval' 有mini的选项吧？

yinjunbo / IS-Fusion

gt_bboxes_2d = info['gt_boxes_2d'][mask] #11

NOTE load image annotations