desh2608 / dover-lap Goto Github PK

View Code? Open in Web Editor NEW

74.0 4.0 13.0 1.03 MB

Python package for combining diarization system outputs.

License: MIT License

Python 100.00%

dover-lap diarization ensemble-machine-learning

dover-lap's People

Contributors

Stargazers

Watchers

Forkers

entn-at gooltz 520liuxing ishine twistedmove damian-romero zaouk normonisping hbredin dtdwind 88aggressive coreycole lefteris12

dover-lap's Issues

Result depends on order of input rttms

Hi,

I made DOVERLap of two RTTMs and has found that the resulting RTTM differs significantly depending of the order of input RTTMs in the command line. Is it expected behaviour of bug? If it is expected, how can I control this process?

Best regards,
Maxim

spyder version

Hello,

Thank you for this project, which has helped many people.

However, I have noticed a minor issue in its usage.

After installing dover-lap with pip install dover-lap, it overwrote my original spyder installation, resulting in a change in the output format of my scores.

This has caused me some inconvenience.

Would it be possible to address this issue?

DIHARD modifications

When ties occur, assign region to all speakers instead of dividing uniformly.
Provide additional argument to supply weights.

Bug: No module named 'dover_lap.libs.metrics'

Python version == 3.6
Installation method: from source
Description:
- After running the example from README.md dover-lap --label-mapping hungarian egs/ami/rttm_dl_test egs/ami/rttm_test_*, the following error message appears:
- dover-lap/dover_lap/src/mapping/hungarian.py", line 9, in <module> from dover_lap.libs.metrics import DER ModuleNotFoundError: No module named 'dover_lap.libs.metrics'
Reason: It looks like DER was removed in a previous refactor
Fix:
- Remove line 9 from ``dover-lap/dover_lap/src/mapping/hungarian.py`

Combining mixed type hypotheses

In its current form, DOVER-Lap does not elegantly handle the case when the input hypotheses are mixed-type. By "mixed type", we mean that some of them have overlapping segments and others do not. It seems the "label mapping" can still be used for this setting, since it uses "relative" overlaps (i.e. intersection-over-union) instead of absolute overlaps to compute speaker similarity. The problem would arise in "label voting", and this needs some new heuristics (or confidence measures) to estimate the actual number of speakers in any region.

label mapping cost_tensor broken in numpy 1.24

Tested that it is working in numpy==1.23.5. Installing numpy==1.24.0 breaks the label mapping step where the cost_tensor is calculated using numpy.sum()

NEP 34 might be helpful?

dover-lap out.rttm egs/ami/rttm_test_*

Loading speaker turns from input RTTMs...
Merging overlapping speaker turns...
Processing file EN2002a.Mix-Headset..
Traceback (most recent call last):
  File "/home/corey/anaconda3/envs/omnicom_env/bin/dover-lap", line 33, in <module>
    sys.exit(load_entry_point('dover-lap', 'console_scripts', 'dover-lap')())
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/click/core.py", line 1157, in __call__
    return self.main(*args, **kwargs)
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/click/core.py", line 1078, in main
    rv = self.invoke(ctx)
  File "/media/data/dev/dover-lap/dover_lap/libs/utils.py", line 86, in invoke
    super(CommandOptionRequiredClass, self).invoke(ctx)
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/click/core.py", line 1434, in invoke
    return ctx.invoke(self.callback, **ctx.params)
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/click/core.py", line 783, in invoke
    return __callback(*args, **kwargs)
  File "/media/data/dev/dover-lap/dover_lap/dover_lap.py", line 164, in main
    file_to_out_turns[file_id] = DOVERLap.combine_turns_list(
  File "/media/data/dev/dover-lap/dover_lap/src/doverlap.py", line 26, in combine_turns_list
    mapped_turns_list, weights = LabelMapping.get_mapped_turns_list(
  File "/media/data/dev/dover-lap/dover_lap/src/label_mapping.py", line 31, in get_mapped_turns_list
    label_mapping, weights = greedy_map.compute_mapping(turns_list)
  File "/media/data/dev/dover-lap/dover_lap/src/mapping/greedy.py", line 25, in compute_mapping
    cost_tensor, pairwise_costs = self.compute_cost_tensor(turns_list)
  File "/media/data/dev/dover-lap/dover_lap/src/mapping/greedy.py", line 95, in compute_cost_tensor
    cost_tensor = np.sum(list(pairwise_costs.values()))
  File "<__array_function__ internals>", line 200, in sum
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 2324, in sum
    return _wrapreduction(a, np.add, 'sum', axis, dtype, out, keepdims=keepdims,
  File "/home/corey/anaconda3/envs/omnicom_env/lib/python3.8/site-packages/numpy/core/fromnumeric.py", line 86, in _wrapreduction
    return ufunc.reduce(obj, axis, dtype, out, **passkwargs)
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (3,) + inhomogeneous part.

Recommend Projects

React

A declarative, efficient, and flexible JavaScript library for building user interfaces.
Vue.js

🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript

TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
TensorFlow

An Open Source Machine Learning Framework for Everyone
Django

The Web framework for perfectionists with deadlines.
Laravel

A PHP framework for web artisans
D3

Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

javascript

JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
web

Some thing interesting about web. New door for the world.
server

A server is a program made to process requests and deliver data to clients.
Machine learning

Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Visualization

Some thing interesting about visualization, use data art
Game

Some thing interesting about game, make everyone happy.

Recommend Org

Facebook

We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft

Open source projects and samples from Microsoft.
Google

Google ❤️ Open Source for everyone.
Alibaba

Alibaba Open Source for everyone
D3

Data-Driven Documents codes.
Tencent

China tencent open source team.