Backward compatible (1.3.0a0 > 1.2.2 to 1.2.6)

[aereimer]

"Greetings! I recently experimented with the alpha version 1.3.0a0 of and tested it by opening a project that I initiated in previous versions (1.2.2 to 1.2.6). I added some additional labeled frames, trained the model, and performed successful inferences. However, I encountered issues when attempting to convert the h5 output to slp and then the slp file to csv using a script by @Toshea111 (#1089 (comment)). I suspect that the incompatibility problem might be related to changes in the file structure of the new version's output. Could you confirm if this issue is due to a bug in the newest version or if the file structure of the new version has indeed changed?"

[talmo]

Hi @aereimer,

Do you mind sharing the specific error you're getting now? That'll help narrow down the cause.

Cheers,

Talmo

[aereimer]

Hi @talmo, sorry for not pasting the log here before. So, here I tried to open the slp project (created initially in 1.2.2, modified in 1.2.6, and finally modified and saved using 1.3.0a0) using versions < 1.3.0a0 versions. The error below appears specifically when I tried to open the project using version 1.2.6 ( the same thing happened with other versions, such as 1.2.9). Just one thing, I did not modified the skeleton or added more bodyparts while using the 1.3.0a0 . Thanks for checking this! Parabéns pelo excelente trabalho!

Traceback (most recent call last):
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\commands.py", line 262, in openProject
    self.execute(OpenProject, filename=filename, first_open=first_open)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\commands.py", line 240, in execute
    command().execute(context=self, params=kwargs)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\commands.py", line 137, in execute
    self.do_with_signal(context, params)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\commands.py", line 161, in do_with_signal
    cls.do_action(context, params)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\commands.py", line 606, in do_action
    context.app.loadProjectFile(filename)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\gui\app.py", line 1451, in loadProjectFile
    labels = Labels.load_file(filename, video_search=gui_video_callback)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\dataset.py", line 1943, in load_file
    filename, for_object="labels", video_search=video_search, *args, **kwargs
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\format\main.py", line 113, in read
    return disp.read(filename, *args, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\format\dispatch.py", line 56, in read
    return adaptor.read(file, *args, **kwargs)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\format\hdf5.py", line 139, in read
    labels = cls.read_headers(file, video_search, match_to)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\format\hdf5.py", line 124, in read_headers
    labels = labels_json.LabelsJsonAdaptor.from_json_data(dicts, match_to=match_to)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\io\format\labels_json.py", line 384, in from_json_data
    dicts["skeletons"], List[Skeleton]
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\cattr\converters.py", line 201, in structure
    return self._structure_func.dispatch(cl)(obj, cl)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\cattr\converters.py", line 336, in _structure_list
    for e in obj
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\cattr\converters.py", line 336, in <listcomp>
    for e in obj
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\skeleton.py", line 248, in <lambda>
    Skeleton, lambda x, cls: Skeleton.from_dict(x, idx_to_node)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\skeleton.py", line 931, in from_dict
    return Skeleton.from_json(json.dumps(d), node_to_idx)
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\sleap\skeleton.py", line 1030, in from_json
    graph = json_graph.node_link_graph(jsonpickle.decode(json_str))
  File "C:\ProgramData\Anaconda3\envs\sleap_develop\lib\site-packages\networkx\readwrite\json_graph\node_link.py", line 166, in node_link_graph
    for d in data["nodes"]:
KeyError: 'nodes'

[Toshea111]

Dear aereimer,

The issue might be that you are trying to convert '.slp' files, when the notebook is designed for converting '.h5' files to '.csv'.

Try directly inputting the '.h5' output, rather than converting it into '.slp' format first.

Best wishes,

Thomas

[aereimer]

Dear @Toshea111,

Thanks for pointing this out. I believe I’ve tried first the slp file because the colab file originally was pointing to a slp file ( [email protected] ). But I also tried the h5 file (exported from SLEAP v1.3.0a0 ) and it did not work, the error message is below - that's also the reason I've asked about changes in the output file in this newer version.

Thanks once again for you input.
Best,
Adriano

`KeyError Traceback (most recent call last)
in
----> 1 labels = sio.load_slp("labels.v001.014_R23_EPM_session_1_20211206_22_adjusted.analysis.h5")
2 # labels = sleap.load_file("sleap1.3")
3 labels

4 frames
/usr/local/lib/python3.9/dist-packages/sleap_io/io/main.py in load_slp(filename)
17 The dataset as a Labels object.
18 """
---> 19 return slp.read_labels(filename)
20
21

/usr/local/lib/python3.9/dist-packages/sleap_io/io/slp.py in read_labels(labels_path)
239 The processed Labels object.
240 """
--> 241 tracks = read_tracks(labels_path)
242 videos = read_videos(labels_path)
243 skeletons = read_skeletons(labels_path)

/usr/local/lib/python3.9/dist-packages/sleap_io/io/slp.py in read_tracks(labels_path)
57 A list of Track objects.
58 """
---> 59 tracks = [json.loads(x) for x in read_hdf5_dataset(labels_path, "tracks_json")]
60 track_objects = []
61 for track in tracks:

/usr/local/lib/python3.9/dist-packages/sleap_io/io/utils.py in read_hdf5_dataset(filename, dataset)
20 """
21 with h5py.File(filename, "r") as f:
---> 22 data = f[dataset][()]
23 return data
24

h5py/_objects.pyx in h5py._objects.with_phil.wrapper()

h5py/_objects.pyx in h5py._objects.with_phil.wrapper()

/usr/local/lib/python3.9/dist-packages/h5py/_hl/group.py in getitem(self, name)
286 raise ValueError("Invalid HDF5 object reference")
287 else:
--> 288 oid = h5o.open(self.id, self._e(name), lapl=self._lapl)
289
290 otype = h5i.get_type(oid)

h5py/_objects.pyx in h5py._objects.with_phil.wrapper()

h5py/_objects.pyx in h5py._objects.with_phil.wrapper()

h5py/h5o.pyx in h5py.h5o.open()

KeyError: "Unable to open object (object 'tracks_json' doesn't exist)"`

[Toshea111]

Dear aereimer,

Looking at the link, I think that is an older version of the notebook.

There is an updated version here:

https://github.com/Toshea111/sleap/blob/develop/docs/notebooks/Convert_HDF5_to_CSV_updated.ipynb

Let me know if the issue persists.

Best wishes,

Thomas

[roomrys]

Hi @aereimer,

I can sheepishly confirm that the file structure has changed. We added two new attributes in the Skeleton class called description and preview_image. Consequently, while you are able to open older files on v1.3.0a0, files from this version are unable to be opened on previous versions - which is not a problem if working on a single SLEAP version. However, I can see how this could easily become extremely annoying if working on multiple workstations which all have different versions of SLEAP installed.

The field we added is currently not used except for provided template skeletons, so we could selectively write the field depending on whether the skeleton is a template (manually specified via API - only by those looking to add to our collection of template skeletons). I opened PR #1222 to rectify this.

Thanks,
Liezl

[aereimer]

Dear @Toshea111,

I have tried the new script that you shared, and it works great! It is compatible with all SLEAP versions that I have tested. I made minor changes to the script to suit my project's requirements (as it has a single track; I still wanted to get the score for each body part/frame), but apart from that, it is perfect. Thank you once again for sharing the notebook!

Best,
Adriano

[aereimer]

Dear @roomrys,

Thanks for getting back to me and for your prompt attention to this matter.

Best,
Adriano

[Toshea111]

Dear Adriano,

Excellent, and good to hear that it works with the latest version.

@roomrys I am happy to incorporate this into a future analysis notebook, if that would be useful?

Best wishes,

Thomas

[roomrys]

Hi @Toshea111,

Yes! This notebook is gaining a lot of traction. We could even convert your notebook into an adaptor for using in the GUI/API/CLI instead of having it as a separate entity. If you start a PR, we can collaborate (not sure if you would be able work on a talmolab/sleap branch, but I can pop to your fork to get the basic structure set-up). We would need to

1. Add an adaptor file to the format folder which implements a subclass of the Adaptor class
   

   sleap/sleap/io/format/adaptor.py

   Lines 20 to 28 in 4bd0171

   	@attr.s(auto_attribs=True)
   	class Adaptor:
   	"""
   	File format adaptor base class.
   	An adaptor handles reading and/or writing a specific file format. To add
   	support for a new file format, you'll create a new class which inherits from
   	the Adaptor base class and implements the relevant functions.
   	"""

2. Connect the adaptor to the GUI (various places)

3. Connect the adaptor to the sleap-convert command
   

   sleap/sleap/io/convert.py

   Lines 100 to 105 in 4bd0171

   	def main(args: list = None):
   	"""Entrypoint for `sleap-convert` CLI for converting .slp to different formats.
   	Args:
   	args: A list of arguments to be passed into sleap-convert.
   	"""

I think copying some of the notebook code into the different methods in the adaptor would be a great start for integration. Let me know what you think. Totally understand if this is more than what you wanted to bite off, and I can take over whatever you'd rather not tackle at the moment.

Thanks,
Liezl

[Toshea111]

Dear Liezl,

I would be happy to collaborate, although I fear that my expertise is limited.

I made the fork 'Toshea111/sleap' a while back, and will invite you as a collaborator.

The notebook is currently located in:

https://github.com/Toshea111/sleap/tree/develop/docs/notebooks

The impetus for the notebook is some downstream software that I am working on, thus further integration would be useful.

Best wishes,

Thomas

[roomrys]

This fix is now available in the latest release v1.3.0. To install:

conda create -y -n sleap -c sleap -c nvidia -c conda-forge sleap=1.3.0

[Toshea111]

Dear Liezl,

Much appreciated.

Best wishes,

Thomas