Could you tell me what is the input for --from path and --to path? is it the whole dataset + annotation file? or is it just annotation file?

When converting from PASCAL to VOC I end up with KITTI annotation files with the label on one line and the annotation info on the following line, and with quote marks in the file. For example:

"
   handgun
  " 0.0 0 -1 1.0 3.0 607.0 266.0 -1 -1 -1 -1 -1 -1 -1

My understanding based on other KITTI files I've seen is that we should instead have everything on one line like so:

handgun 0.0 0 -1 1.0 3.0 607.0 266.0 -1 -1 -1 -1 -1 -1 -1

I assume that this is happening in kitti.py in the KITTIEgestor.egest function.

Am I the only one to have trouble with this? It seems odd that this issue wouldn't come up for others before now.

Hello,
I tried to run the converter, but it fails due to the lack of train.txt, I didn't find it on KITTiI website.

I am working on Stanford drone dataset. I have converted the dataset to KITTI format using these scripts.
https://github.com/rockkingjy/DataFormat_sdd2kitti After conversion, I tried conversion from kitti format to VOC, but I am constantly getting this error.
`Traceback (most recent call last):
File "vod_converter/main.py", line 92, in
filter_images_without_labels=args.filter_images_without_labels))
File "vod_converter/main.py", line 40, in main
filter_images_without_labels=filter_images_without_labels)
File "/home/pranav/vod_converter/vod_converter/converter.py", line 134, in convert
validate_image_detections(image_detections)
File "/home/pranav/vod_converter/vod_converter/converter.py", line 153, in validate_image_detections
raise ValueError(f"Image {image} has out of bounds bounding box {detection}")
ValueError: Image {'id': '1758', 'path': '/home/pranav/test_training/annotation0/training/image_2/1758.png', 'segmented_path': None, 'width': 1424, 'height': 1088} has out of bounds bounding box {'label': 'Pedestrian', 'left': 231.0, 'right': 282.0, 'top': 1078.0, 'bottom': 1088.0}

`
Can you please guide me through this? Why am I getting this error and how to solve it.

When I run code as below,I got :
python3.6 vod_converter/main.py --from udacity-crowdai --from-path home/ycyer/Udatasets1 --to voc --to-path home/ycyer/VOCdevkit/voc
INFO:root:Namespace(filter_images_without_labels=False, from_key='udacity-crowdai', from_path='home/ycyer/Udatasets1', select_only_known_labels=False, to_key='voc', to_path='home/ycyer/VOCdevkit/voc')
Failed to convert from udacity-crowdai to voc: Expected to find home/ycyer/Udatasets1/labels.csv

But I checked the Udatasets1,and found labels.csv is init,can anybody tell me where I was wrong?

what is the problem??
Failed to convert from from kitti to voc: Expected subdirectory training/image_2 within datasets/mydata-kitti.

Hi, I have some troubles trying to run the code for transform from Kitti Dataset format to VOC format with the next log:

$ python vod_converter/main.py --from kitti --from-path /g/data_object_image_2/ --to voc --to-path ./out INFO:root:Namespace(filter_images_without_labels=False, from_key='kitti', from_path='G:/data_object_image_2/', select_only_known_labels=False, to_key='voc', to_path='./out') Traceback (most recent call last): File "vod_converter/main.py", line 92, in <module> filter_images_without_labels=args.filter_images_without_labels)) File "vod_converter/main.py", line 40, in main filter_images_without_labels=filter_images_without_labels) File "C:\Users\ASDF\Desktop\GPUexp\vod-converter\vod_converter\converter.py", line 140, in convert egestor.egest(image_detections=image_detections, root=to_path) File "C:\Users\ASDF\Desktop\GPUexp\vod-converter\vod_converter\voc.py", line 122, in egest shutil.copyfile(image['path'], "%s/%d.%s" % (images_path, image_id, src_extension)) TypeError: %d format: a number is required, not str (py36)

Later if I modify this "%s/%d.%s" % (images_path, image_id, src_extension)) to ------>{}/{}.{}" .format(images_path, image_id,src_extension)` format, the next error message is show:

`
$ python vod_converter/main.py --from kitti --from-path /g/data_object_image_2/ --to voc --to-path ./out
INFO:root:Namespace(filter_images_without_labels=False, from_key='kitti', from_path='G:/data_object_image_2/', select_only_known_labels=False, to_key='voc', to_path='./out')
Traceback (most recent call last):
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 1066, in _escape_cdata
if "&" in text:
TypeError: argument of type 'int' is not iterable

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "vod_converter/main.py", line 92, in
filter_images_without_labels=args.filter_images_without_labels))
File "vod_converter/main.py", line 40, in main
filter_images_without_labels=filter_images_without_labels)
File "C:\Users\ASDF\Desktop\GPUexp\vod-converter\vod_converter\converter.py", line 140, in convert
egestor.egest(image_detections=image_detections, root=to_path)
File "C:\Users\ASDF\Desktop\GPUexp\vod-converter\vod_converter\voc.py", line 164, in egest
ET.ElementTree(xml_root).write("{}/{}.xml".format(str(annotations_path), str(image_id)))
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 776, in write
short_empty_elements=short_empty_elements)
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 941, in _serialize_xml
short_empty_elements=short_empty_elements)
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 941, in _serialize_xml
short_empty_elements=short_empty_elements)
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 938, in _serialize_xml
write(_escape_cdata(text))
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 1074, in _escape_cdata
_raise_serialization_error(text)
File "C:\Users\ASDF.conda\envs\py36\lib\xml\etree\ElementTree.py", line 1057, in _raise_serialization_error
"cannot serialize %r (type %s)" % (text, type(text).name)
TypeError: cannot serialize 3 (type int)
`

Maybe is the python version? although the same happens in py27 fork version of the code, thanks for the feedback

I would like to convert udacity to voc. I wonder whether the converter I am using is right or not because it keep giving me error

First i tried this main.py

~/Documents/vod-converter$ python3.6 vod_converter/main.py --from udacity --from-path datasets/mydata-udacity/ --to voc --to-path datasets/mydata-voc/

output is

Traceback (most recent call last):
  File "vod_converter/main.py", line 12, in <module>
    import kitti
  File "/home/borinyun/Documents/vod-converter/vod_converter/kitti.py", line 35, in <module>
    from PIL import Image
  File "/usr/lib/python3/dist-packages/PIL/Image.py", line 66, in <module>
    from PIL import _imaging as core
ImportError: cannot import name '_imaging'
Error in sys.excepthook:
Traceback (most recent call last):
  File "/usr/lib/python3/dist-packages/apport_python_hook.py", line 63, in apport_excepthook
    from apport.fileutils import likely_packaged, get_recent_crashes
  File "/usr/lib/python3/dist-packages/apport/__init__.py", line 5, in <module>
    from apport.report import Report
  File "/usr/lib/python3/dist-packages/apport/report.py", line 30, in <module>
    import apport.fileutils
  File "/usr/lib/python3/dist-packages/apport/fileutils.py", line 23, in <module>
    from apport.packaging_impl import impl as packaging
  File "/usr/lib/python3/dist-packages/apport/packaging_impl.py", line 23, in <module>
    import apt
  File "/usr/lib/python3/dist-packages/apt/__init__.py", line 23, in <module>
    import apt_pkg
ModuleNotFoundError: No module named 'apt_pkg'

Original exception was:
Traceback (most recent call last):
  File "vod_converter/main.py", line 12, in <module>
    import kitti
  File "/home/borinyun/Documents/vod-converter/vod_converter/kitti.py", line 35, in <module>
    from PIL import Image
  File "/usr/lib/python3/dist-packages/PIL/Image.py", line 66, in <module>
    from PIL import _imaging as core
ImportError: cannot import name '_imaging'

Then I tried this udacity.py

~/Documents/vod-converter$ python3.6 vod_converter/udacity.py --from udacity --from-path datasets/mydata-udacity/ --to voc --to-path datasets/mydata-voc/

output

Traceback (most recent call last):
  File "vod_converter/udacity.py", line 10, in <module>
    from PIL import Image
  File "/usr/lib/python3/dist-packages/PIL/Image.py", line 66, in <module>
    from PIL import _imaging as core
ImportError: cannot import name '_imaging'
Error in sys.excepthook:
Traceback (most recent call last):
  File "/usr/lib/python3/dist-packages/apport_python_hook.py", line 63, in apport_excepthook
    from apport.fileutils import likely_packaged, get_recent_crashes
  File "/usr/lib/python3/dist-packages/apport/__init__.py", line 5, in <module>
    from apport.report import Report
  File "/usr/lib/python3/dist-packages/apport/report.py", line 30, in <module>
    import apport.fileutils
  File "/usr/lib/python3/dist-packages/apport/fileutils.py", line 23, in <module>
    from apport.packaging_impl import impl as packaging
  File "/usr/lib/python3/dist-packages/apport/packaging_impl.py", line 23, in <module>
    import apt
  File "/usr/lib/python3/dist-packages/apt/__init__.py", line 23, in <module>
    import apt_pkg
ModuleNotFoundError: No module named 'apt_pkg'

Original exception was:
Traceback (most recent call last):
  File "vod_converter/udacity.py", line 10, in <module>
    from PIL import Image
  File "/usr/lib/python3/dist-packages/PIL/Image.py", line 66, in <module>
    from PIL import _imaging as core
ImportError: cannot import name '_imaging'

Hi
I am trying to convert txt to pascal voc, but I couldn't I tried python2 and puthon3.6, both give same error

(py3.6) abdu@grasp-uoit:~/1vod-converter$ python3.6 vod_converter/main.py --from kitti --from-path /home/abdu/datasets/mydata-kitti --to voc --to-path /home/abdu/datasets/mydata-voc
INFO:root:Namespace(filter_images_without_labels=False, from_key='kitti', from_path='/home/abdu/datasets/mydata-kitti', select_only_known_labels=False, to_key='voc', to_path='/home/abdu/datasets/mydata-voc')
Failed to convert from kitti to voc: Expected subdirectory training/image_2 within /home/abdu/datasets/mydata-kitti

Any suggestions please ???

I made a tool to convert all f-strings inside the vod-converter to older python strings
https://github.com/supernlogn/f-string-deconverter

So the project can be compatible to older python3 versions.

Why do the labels not contain 3d bounding box coordinates? They seem to be replaced by -1

Both KITTI and VOC have a notion of occlusion, it should be preserved per discussion in #13

python main.py --from udacity-crowdai --from-path ../datasets/crowdai --to voc --to-path ../datasets/crowdai-voc
INFO:root:Namespace(filter_images_without_labels=False, from_key='udacity-crowdai', from_path='../datasets/crowdai', select_only_known_labels=False, to_key='voc', to_path='../datasets/crowdai-voc')
Successfully converted from udacity-crowdai to voc.

but the folder path "datasets/autti-voc/VOC2012/Annotations" is empty and there is no data

In order to register and upload this to PyPI there should be a setup.py etc. in order to facilitate the process, as described here.

The short version:

1. Get the setup.py file in shape as described here
2. $ rm -rf dist
3. $ python setup.py sdist bdist_wheel
4. $ twine check dist/*
5. $ twine upload --repository-url https://test.pypi.org/legacy/ dist/*
6. $ twine upload dist/*

If this is useful for the project I can make the updates and submit a PR but I don't think I have permission to upload the package to PyPI.

I didn't find the format of the test.txt on the KITTI website. Could anybody put it there?

Thanks!

Traceback (most recent call last):
File "vod_converter/main.py", line 92, in
filter_images_without_labels=args.filter_images_without_labels))
File "vod_converter/main.py", line 40, in main
filter_images_without_labels=filter_images_without_labels)
File "/media/liufan/689097F39097C64A/ubuntu16.04/dataConverter/vod-converter/vod_converter/converter.py", line 133, in convert
image_detections = ingestor.ingest(from_path)
File "/media/liufan/689097F39097C64A/ubuntu16.04/dataConverter/vod-converter/vod_converter/kitti.py", line 60, in ingest
return [self._get_image_detection(path, image_name, image_ext=image_ext) for image_name in image_ids]
File "/media/liufan/689097F39097C64A/ubuntu16.04/dataConverter/vod-converter/vod_converter/kitti.py", line 60, in
return [self._get_image_detection(path, image_name, image_ext=image_ext) for image_name in image_ids]
File "/media/liufan/689097F39097C64A/ubuntu16.04/dataConverter/vod-converter/vod_converter/kitti.py", line 75, in _get_image_detection
detections = self._get_detections(detections_fpath)
File "/media/liufan/689097F39097C64A/ubuntu16.04/dataConverter/vod-converter/vod_converter/kitti.py", line 95, in _get_detections
x1, y1, x2, y2 = map(float, row[4:8])

Content format of train.txt from KITTI consists of 2 things: path to image file and a path to label file of that image. Even though this piece of data may seem redundant, I think it is better to have a consistent format with KITTI.

Fixed in #1

Hi. I am trying to convert annotation files from pascal VOC to Kitti format. When I run the command,
python3.6 vod_converter/main.py --from voc --from-path ~/Desktop/voc --to kitti --to-path ~/Desktop/kitti
I get the error below:

Failed to convert from voc to kitti: Expected subdirectory ImageSets within /Users/sowmya/Desktop/voc/VOC2012

Could you please let me know where the conversion possibly failed? Also, it would be really helpful to specify the dependencies in ReadMe, such as PIL or jsonschema libraries.

Thank you so much,
Sowmya

@krosaen Question: Do you intent make this repo compatible with other python versions?

If yes, I can make a PR for that.

do you convert to any intermediate representation first before converting to other formats?

I am trying to convert this Kitti Road/Lane Detection Evaluation 2013 to Pascal voc like dataset.

I have renamed calib folders (found under training and testing folders) to label_2 and I have created the train.txt like it was instructed. The first lines of train.txt are the following:
um_000000
um_000001
um_000002
um_000003
um_000004
um_000005
um_000006
um_000007

When I convert it I get the following validation error:

INFO:root:Namespace(filter_images_without_labels=False, from_key='kitti', from_path='./Kitti/data_road', select_only_known_labels=False, to_key='voc', to_path='./Kitti-voc')
Traceback (most recent call last):
File "/home/bm-pc/Documents/Thesis/vod-converter-master/vod_converter/converter.py", line 147, in validate_image_detections
validate_schema(image_detection, IMAGE_DETECTION_SCHEMA)
File "/home/bm-pc/Documents/Thesis/vod-converter-master/vod_converter/converter.py", line 22, in validate_schema
return raw_validate(data, schema, types={"array": (list, tuple)})
File "/home/bm-pc/.local/lib/python3.6/site-packages/jsonschema/validators.py", line 541, in validate
cls(schema, *args, **kwargs).validate(instance)
File "/home/bm-pc/.local/lib/python3.6/site-packages/jsonschema/validators.py", line 130, in validate
raise error
jsonschema.exceptions.ValidationError: -387.5744 is less than the minimum of 0

Failed validating 'minimum' in schema['properties']['detections']['items']['properties']['left']:
{'minimum': 0, 'type': 'number'}

On instance['detections'][1]['left']:
-387.5744

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "../vod-converter-master/vod_converter/main.py", line 92, in
filter_images_without_labels=args.filter_images_without_labels))
File "../vod-converter-master/vod_converter/main.py", line 40, in main
filter_images_without_labels=filter_images_without_labels)
File "/home/bm-pc/Documents/Thesis/vod-converter-master/vod_converter/converter.py", line 134, in convert
validate_image_detections(image_detections)
File "/home/bm-pc/Documents/Thesis/vod-converter-master/vod_converter/converter.py", line 149, in validate_image_detections
raise Exception(f"at index {i}") from se
Exception: at index 0

Could you tell me please what did I miss or this converter is intended to be used only with these Kitti datasets:

• KITTI
• KITTI tracking

Thank you.

When the VOCIngestor._get_detection function parses bounding box coordinate values it decrements the values. Why is this so?

For example, if my image is 200 pixels wide and contains a box that's full width with xmin==0 and xmax==199 then this code results in a detection with left==-1 and right==198. Is this really what we want? It gives me trouble in this situation, in that it results in invalid boxes since -1 is out-of-range.

Also the box coordinate decrementations result in the box being pushed to the left/top by one pixel in every case. Is this really what we want? Perhaps there was some justification for the decrementations when the code was written that's just not obvious to me?

when I run the code :
python3.6 vod_converter/main.py --from udacity --from-/home/ycyer/Udataset1 --to voc --to-path /home/ycyer/VOCdevkit/VOC2012/JPEGImages2
the output is :
sage: main.py [-h] --from FROM_KEY --from-path FROM_PATH --to TO_KEY
--to-path TO_PATH [--select-only-known-labels]
[--filter-images-without-labels]
main.py: error: the following arguments are required: --from-path

I think I didn't write the paty right,but actually I don't know how to set the voc path, can anyone tell me how to fix this?? Thanks for advance!!!

How does each class in Kitti tracking are converted to Voc?

'Car', 'Van', 'Truck', 'Pedestrian', 'Person_sitting', 'Cyclist', 'Tram', 'Misc' and 'DontCare'

I'm trying to convert from udacity-autti and udacity-crowdai to voc, but only the raw folders are generated without any files inside. Converting from KITTI to voc works like a charm.

I am using this command in Anaconda:

(C:\ProgramData\Anaconda3) D:\data\KITTI_dataset\KITTI_2_VOC_converter_vod-converter-master\vod-converter-master>python vod_converter/main.py --from udacity-crowdai --from-path D:\data\CrowdAI_dataset\object-detection-crowdai\ --to voc --to-path D:\data\CrowdAI_voc\object-detection-crowdai\

Getting this output:

INFO:root:Namespace(filter_images_without_labels=False, from_key='udacity-crowdai', from_path='D:\data\CrowdAI_dataset\object-detection-crowdai\', select_only_known_labels=False, to_key='voc', to_path='D:\data\CrowdAI_voc\object-detection-crowdai\')
Successfully converted from udacity-crowdai to voc.

This folder is generated:
_D:\data\CrowdAI_voc\object-detection-crowdai\VOC2012_
with subfolders:
Annotations
ImageSets\Main
JPEGImages

same for
--from udacity-autti

Images and labels.csv are placed in D:\data\CrowdAI_dataset\object-detection-crowdai

what am I doing wrong or is there a problem with the scripts?

The annotations files has some mistakes after the conversion.

For example :
https://github.com/umautobots/vod-converter/blob/master/vod_converter/kitti.py#L100
in this line the bounding box right border is limited to 1023, although kitti images size is much larger than 1023.

@krosaen I am trying to convert from voc to kitti

home@home-lnx:~/Desktop/vod-converter-master$ python3.6 vod_converter/main.py --from voc --from-path datasets/ --to kitti --to-path dataset
INFO:root:Namespace(filter_images_without_labels=False, from_key='voc', from_path='datasets/', select_only_known_labels=False, to_key='kitti', to_path='dataset')
Failed to convert from voc to kitti: Expected subdirectory ImageSets within datasets//VOC2012