`pytorch3d.io().load_mesh()` can not handle `.off` file without newline or space after 'OFF'

[yadiszzz]

🐛 Bugs / Unexpected behaviors

Lots of off files from ModelNet40 dataset don't have the newline or space after 'OFF' at the beginning of the file.

e.g. for file monitor/train/monitor_0284.off

It begins with

OFF1046 900 0
-0.797495 -6.623700 9.540650 
...

instead of

OFF
1046 900 0
-0.797495 -6.623700 9.540650 
...

But the header skip logic below can not handle it

pytorch3d/pytorch3d/io/off_io.py

Lines 229 to 237 in 097b0ef

	if header.lower() in (b"off\n", b"off\r\n", "off\n"):
	header = file.readline()
	while _is_line_empty(header):
	header = file.readline()
	items = header.split(b" ")
	if len(items) and items[0].lower() in ("off", b"off"):
	items = items[1:]

Instructions To Reproduce the Issue:

Please include the following (depending on what the issue is):

1. Any changes you made (git diff) or code you wrote

from pytorch3d.io import IO

mesh = IO().load_mesh(path_to_off_file)

2. The exact command(s) you ran:

In order to read off file, I installed the pytorch3d from source via

pip install "git+https://github.com/facebookresearch/pytorch3d.git"

Since the last stable 0.4.0 doesn't contain 0345f86.

3. What you observed (including the full logs):

Traceback (most recent call last):                                                                                                                                                                                                                                              
  File "<...>/env/lib/python3.8/site-packages/pytorch3d/io/off_io.py", line 242, in _load_off_stream                                                                                                                         
    n_verts = int(items[0])                                                                                                                                                                                                                                                     
ValueError: invalid literal for int() with base 10: b'OFF150'                                                                                                                                                                                                                   
                                                                                                                                                                                                                                                                                
During handling of the above exception, another exception occurred: 

Traceback (most recent call last):                                                                                                                                                                                                                                              

[omitted stack]
  File "<my_project>/src/datamodules/datasets/modelnet40_dataset.py", line 58, in off_vertex_parser
    mesh = IO().load_mesh(path_to_off_file)
  File "/data/daizhu/workspace/pc_mesh_transformer_pylight/conda/env/lib/python3.8/site-packages/pytorch3d/io/pluggable.py", line 124, in load_mesh
    mesh = mesh_interpreter.read(
  File "/data/daizhu/workspace/pc_mesh_transformer_pylight/conda/env/lib/python3.8/site-packages/pytorch3d/io/off_io.py", line 430, in read
    data = _load_off_stream(f)
  File "/data/daizhu/workspace/pc_mesh_transformer_pylight/conda/env/lib/python3.8/site-packages/pytorch3d/io/off_io.py", line 244, in _load_off_stream
    raise ValueError("Invalid counts line: %s" % header)
ValueError: Invalid counts line: b'OFF150 92 0\n'

The above exception was the direct cause of the following exception:
  File "./run.py", line 31, in <module>
    main()
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/main.py", line 49, in decorated_main
    _run_hydra(
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/_internal/utils.py", line 370, in _run_hydra
    run_and_report(
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/_internal/utils.py", line 215, in run_and_report
    raise ex
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/_internal/utils.py", line 212, in run_and_report
    return func()
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/_internal/utils.py", line 371, in <lambda>
    lambda: hydra.run(
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/_internal/hydra.py", line 107, in run
    _ = ret.return_value
  File "<my_project>/conda/env/lib/python3.8/site-packages/hydra/core/utils.py", line 207, in return_value
    raise HydraJobException(
hydra.errors.HydraJobException: Error executing job with overrides: []

[yadiszzz]

I'm happy to send a pull request if this is the behavior should be supported.

[bottler]

I think it should be fixed, and we can do it internally. I suggest that the lines (around line 235)

    items = header.split(b" ")
    if len(items) and items[0].lower() in ("off", b"off"):
        items = items[1:]

are replaced with

    if header[:3].lower() == b"off":
        header = header[3:]
    items = header.split(b" ")

It would be good if you could confirm that this makes it work.

[bottler]

And thank you for reporting this thoroughly!

[yadiszzz]

I think it should be fixed, and we can do it internally. I suggest that the lines (around line 235)

    items = header.split(b" ")
    if len(items) and items[0].lower() in ("off", b"off"):
        items = items[1:]

are replaced with

    if header[:3].lower() == b"off":
        header = header[3:]
    items = header.split(b" ")

It would be good if you could confirm that this makes it work.

It works.

And it is better to merge with the check above as well

 if header.lower() in (b"off\n", b"off\r\n", "off\n"): 
     header = file.readline() 

I put the change together in a pull request, PTAL.