Files
pyannote-audio/tutorials/overlapped_speech_detection.ipynb

621 lines
147 KiB
Plaintext
Raw Permalink Blame History

This file contains invisible Unicode characters
This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"<a href=\"https://colab.research.google.com/github/pyannote/pyannote-audio/blob/develop/tutorials/overlapped_speech_detection.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"# Finetuning a segmentation model on an overlapped speech detection task with `pyannote.audio`\n",
"\n",
"Overlapped speech detection (OSD) is the task of detecting regions where at least two speakers are speaking at the same time. In this notebook, we will finetune a segmentation model on the OSD task, then evaluate an OSD pipeline on AMI database."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Tutorial setup"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### `Google Colab` setup"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If you are running this tutorial on `Colab`, execute the following commands in order to setup `Colab` environment. These commands will install `pyannote.audio` and download a mini version of the `AMI` corpus."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"!pip install -qq pyannote.audio==3.1.1\n",
"!pip install -qq ipython==7.34.0\n",
"!git clone https://github.com/pyannote/AMI-diarization-setup.git\n",
"%cd ./AMI-diarization-setup/pyannote/\n",
"!bash ./download_ami_mini.sh\n",
"%cd /content"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"⚠ Restart the runtime (Runtime > Restart session)."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Non `Google Colab` setup"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"If you are not using `Colab`, this tutorial assumes that\n",
"* `pyannote.audio` has been installed\n",
"* the [AMI corpus](https://groups.inf.ed.ac.uk/ami/corpus/) has already been [setup for use with `pyannote`](https://github.com/pyannote/AMI-diarization-setup/tree/main/pyannote)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Protocol"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Firstly, we define a protocol, here `AMI.SpeakerDiarization.mini`"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.database import registry, FileFinder\n",
"\n",
"registry.load_database(\"AMI-diarization-setup/pyannote/database.yml\")\n",
"protocol = registry.get_protocol(\"AMI.SpeakerDiarization.mini\", preprocessors={\"audio\": FileFinder()})"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"`pyannote.database` *protocols* usually define \n",
"* a training set: `for training_file in protocol.train(): ...`, \n",
"* a validation set: `for validation_file in protocol.development(): ...` \n",
"* an evaluation set `for evaluation_file in protocol.test(): ...`"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Let's listen to the first training file and visualize its reference annotation:"
]
},
{
"cell_type": "code",
"execution_count": 24,
"metadata": {},
"outputs": [],
"source": [
"first_training_file = next(protocol.train())"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.audio.utils.preview import listen\n",
"listen(first_training_file)"
]
},
{
"cell_type": "code",
"execution_count": 26,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<pyannote.core.annotation.Annotation at 0x7f65965dc4c0>"
]
},
"execution_count": 26,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"first_training_file['annotation']"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"The expected output of a perfect overlapped speech detection pipeline would look like this:"
]
},
{
"cell_type": "code",
"execution_count": 27,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Timeline(uri=IS1000a, segments=[<Segment(96.02, 97.48)>, <Segment(114.49, 114.73)>, <Segment(126.4, 126.88)>, <Segment(133.07, 133.43)>, <Segment(133.51, 134.12)>, <Segment(144.85, 145.79)>, <Segment(165.37, 165.59)>, <Segment(166.09, 166.72)>, <Segment(170.19, 170.87)>, <Segment(171.11, 171.13)>, <Segment(188.04, 188.8)>, <Segment(213.26, 213.44)>, <Segment(216.64, 216.87)>, <Segment(238.76, 239.02)>, <Segment(298.36, 298.6)>, <Segment(299.33, 299.98)>, <Segment(442.8, 443.39)>, <Segment(478.02, 478.41)>, <Segment(521.19, 521.31)>, <Segment(522.47, 524.08)>, <Segment(524.13, 524.93)>, <Segment(527.85, 528.7)>, <Segment(531.12, 532.15)>, <Segment(533.24, 533.85)>, <Segment(534.97, 535.26)>, <Segment(555.66, 556.98)>, <Segment(562.6, 563.09)>, <Segment(570.62, 571.23)>, <Segment(580.98, 581.59)>, <Segment(595.69, 597.32)>, <Segment(617.72, 618.37)>, <Segment(641.35, 641.68)>, <Segment(688.96, 691.36)>, <Segment(691.38, 691.81)>, <Segment(692.26, 692.78)>, <Segment(694.35, 694.56)>, <Segment(695.69, 696.81)>, <Segment(697.88, 698.78)>, <Segment(698.95, 699.68)>, <Segment(700.6, 701.48)>, <Segment(703.49, 706.66)>, <Segment(710.34, 710.67)>, <Segment(757.81, 761.93)>, <Segment(768.01, 768.7)>, <Segment(798.05, 798.77)>, <Segment(809.98, 810.24)>, <Segment(813.7, 813.98)>, <Segment(815.64, 815.66)>, <Segment(820.64, 821.18)>, <Segment(912.96, 912.99)>, <Segment(922.14, 922.67)>, <Segment(923.86, 925.43)>, <Segment(932.51, 934.92)>, <Segment(956.47, 957.21)>, <Segment(971.35, 971.55)>, <Segment(976.13, 977.09)>, <Segment(977.43, 977.65)>, <Segment(980, 980.37)>, <Segment(980.64, 980.69)>, <Segment(982.85, 983.27)>, <Segment(994.15, 994.17)>, <Segment(1001.3, 1001.57)>, <Segment(1004.83, 1005.38)>, <Segment(1010.87, 1011.43)>, <Segment(1013.07, 1015.02)>, <Segment(1017.6, 1017.87)>, <Segment(1023.89, 1024.62)>, <Segment(1028.48, 1028.84)>, <Segment(1029.29, 1030.96)>, <Segment(1033.04, 1034.18)>, <Segment(1035.93, 1036.8)>, <Segment(1059.18, 1059.92)>, <Segment(1064.39, 1064.54)>, <Segment(1074.05, 1074.54)>, <Segment(1074.99, 1075.36)>, <Segment(1107.91, 1108.98)>, <Segment(1124, 1125.07)>, <Segment(1127.6, 1128.36)>, <Segment(1133.83, 1134.32)>, <Segment(1134.45, 1136.06)>, <Segment(1136.4, 1136.65)>, <Segment(1138.14, 1138.68)>, <Segment(1141.94, 1142.66)>, <Segment(1143.36, 1143.65)>, <Segment(1145.19, 1146.7)>, <Segment(1149.19, 1149.39)>, <Segment(1190.32, 1190.53)>, <Segment(1202.62, 1202.9)>, <Segment(1204.59, 1204.7)>, <Segment(1208.4, 1208.67)>, <Segment(1212.05, 1212.08)>, <Segment(1215.71, 1215.92)>, <Segment(1217.97, 1218.14)>, <Segment(1218.97, 1219.3)>, <Segment(1225.98, 1226.73)>, <Segment(1229.4, 1229.52)>, <Segment(1230.85, 1231.38)>, <Segment(1233.47, 1233.73)>, <Segment(1235.06, 1235.21)>, <Segment(1236.19, 1236.7)>, <Segment(1237.12, 1237.39)>, <Segment(1242.12, 1243.06)>, <Segment(1244.69, 1244.81)>, <Segment(1247.48, 1247.75)>, <Segment(1248.08, 1248.21)>, <Segment(1249.23, 1249.46)>, <Segment(1251.08, 1251.43)>, <Segment(1254.58, 1254.81)>, <Segment(1254.96, 1255.38)>, <Segment(1255.69, 1256.04)>, <Segment(1257.17, 1264.61)>, <Segment(1266.51, 1269.01)>, <Segment(1272, 1273.63)>, <Segment(1277.07, 1277.8)>, <Segment(1279.1, 1279.93)>, <Segment(1291.39, 1294)>, <Segment(1303.25, 1304.08)>, <Segment(1313.28, 1313.86)>, <Segment(1313.96, 1314.55)>, <Segment(1316.1, 1317.71)>, <Segment(1323.92, 1324.16)>, <Segment(1326.19, 1326.49)>, <Segment(1329.71, 1329.93)>, <Segment(1336.23, 1337.52)>, <Segment(1341.51, 1343.21)>, <Segment(1357.26, 1358.08)>, <Segment(1358.75, 1359.13)>, <Segment(1359.3, 1359.45)>, <Segment(1361.42, 1361.6)>, <Segment(1363.57, 1363.76)>, <Segment(1367.4, 1367.58)>, <Segment(1379.07, 1379.69)>, <Segment(1391.69, 1391.93)>, <Segment(1396.77, 1396.97)>, <Segment(1410.62, 1410.81)>, <Segment(1413.11, 1414.97)>, <Segment(1416.04, 1420.13)>, <Segment(1420.5, 1420.68)>, <Segment(1425.52, 1426.75)>, <Segment(1428.76, 1429.13)>, <Segment(1443.7, 1443.9)>, <Segment(1445.31, 1445.75)>, <Segment(1448.28, 1448.56)>, <Segment(1448.68, 1448.89)>, <Segment(1451.87, 1452.14)>, <Segment(1469.76, 1470.2)>, <Segment(1486.87, 1487.15)>, <Segment(1491.36, 1491.79)>, <Segment(1492.16, 1492.29)>, <Segment(1505.63, 1505.84)>, <Segment(1510, 1510.03)>, <Segment(1511.51, 1511.99)>])>"
]
},
"execution_count": 27,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from pyannote.audio.pipelines.overlapped_speech_detection import OracleOverlappedSpeechDetection\n",
"oracle_osd = OracleOverlappedSpeechDetection()\n",
"\n",
"oracle_osd(first_training_file).get_timeline()"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"## Finetuning\n",
"\n",
"In this part, we propose to finetune the [hf.co/pyannote/segmentation-3.0](https://hf.co/pyannote/segmentation-3.0) on an OSD *task*"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Firstly, we initialize an OSD *task* that describes how the model will be trained:\n",
"\n",
"* `protocol` indicates that we will use files available in `protocol.train()`.\n",
"* `duration=2.` and `batch_size=16` indicates that the model will ingest batches of 16 two seconds long audio chunks."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.audio.tasks import OverlappedSpeechDetection\n",
"osd = OverlappedSpeechDetection(protocol, duration=2., batch_size=16)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The next step is to load the segmentation model. Official [pyannote.audio](https://github.com/pyannote/pyannote-audio) models (i.e. those under the [`pyannote` organization](https://hf.co/pyannote) umbrella) are open-source, but gated. It means that you have to first accept users conditions on their respective Huggingface page to access the pretrained weights and hyper-parameters. Despite this initial process, those models can perfectly be downloaded for later offline use: keep reading this tutorial until the end to learn how to do that.\n",
"\n",
"For instance, to load the speaker segmentation model used in this tutorial, you have to visit [hf.co/pyannote/segmentation-3.0](https://hf.co/pyannote/segmentation-3.0), accept the terms, and log in using `notebook_login` below:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"from huggingface_hub import notebook_login\n",
"notebook_login()"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"We can then initialize the *pretrained model*: it needs to know about the task (`task=osd`) for which it is being trained for:"
]
},
{
"cell_type": "code",
"execution_count": 30,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.audio.core.model import Model\n",
"\n",
"pretrained_model = Model.from_pretrained(\"pyannote/segmentation-3.0\", token=True)\n",
"# we assign the OSD task to the model\n",
"pretrained_model.task = osd"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Now that everything is ready, let's train with `pytorch-lightning`!"
]
},
{
"cell_type": "code",
"execution_count": 33,
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"GPU available: False, used: False\n",
"TPU available: False, using: 0 TPU cores\n",
"IPU available: False, using: 0 IPUs\n",
"HPU available: False, using: 0 HPUs\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"\n",
" | Name | Type | Params | In sizes | Out sizes \n",
"---------------------------------------------------------------------------------------------------------------------\n",
"0 | sincnet | SincNet | 42.6 K | [1, 1, 32000] | [1, 60, 115] \n",
"1 | lstm | LSTM | 1.4 M | [1, 115, 60] | [[1, 115, 256], [[8, 1, 128], [8, 1, 128]]]\n",
"2 | linear | ModuleList | 49.4 K | ? | ? \n",
"3 | classifier | Linear | 129 | [1, 115, 128] | [1, 115, 1] \n",
"4 | activation | Sigmoid | 0 | [1, 115, 1] | [1, 115, 1] \n",
"5 | validation_metric | MetricCollection | 0 | ? | ? \n",
"---------------------------------------------------------------------------------------------------------------------\n",
"1.5 M Trainable params\n",
"0 Non-trainable params\n",
"1.5 M Total params\n",
"5.890 Total estimated model params size (MB)\n"
]
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "2433ca383e8f4470adae1e34ffa53b2f",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Sanity Checking: | | 0/? [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "e1796ee38c0b4587a6d9d957d24c02c1",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Training: | | 0/? [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "0053532719c84b8493e900509347cbf7",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Validation: | | 0/? [00:00<?, ?it/s]"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"`Trainer.fit` stopped: `max_epochs=1` reached.\n"
]
}
],
"source": [
"import pytorch_lightning as pl\n",
"\n",
"trainer = pl.Trainer(max_epochs=1)\n",
"trainer.fit(pretrained_model)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"For the purpose of this tutorial, the model is trained for only 1 epoch. One can obviously expect better performance by training longer and on more data.\n",
"\n",
"See [`pytorch-lightning`](https://www.pytorchlightning.ai/) documentation to learn more about the [`Trainer` API](https://pytorch-lightning.readthedocs.io/en/latest/common/trainer.html), in particular."
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"## Inference\n",
"\n",
"Once trained, we will apply the model on a test file:"
]
},
{
"cell_type": "code",
"execution_count": 34,
"metadata": {},
"outputs": [],
"source": [
"test_file = next(protocol.test())\n",
"# here we use a test file provided by the protocol, but it could be any audio file\n",
"# e.g. test_file = \"/path/to/test.wav\".\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The expected output is the following:"
]
},
{
"cell_type": "code",
"execution_count": 35,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Timeline(uri=IS1009a, segments=[<Segment(55.7, 56.03)>, <Segment(56.9, 57.04)>, <Segment(57.25, 57.7)>, <Segment(62, 62.57)>, <Segment(66.63, 67.29)>, <Segment(76.08, 76.43)>, <Segment(131.13, 131.49)>, <Segment(145.55, 145.79)>, <Segment(159.51, 159.78)>, <Segment(220.46, 221.49)>, <Segment(225.98, 226.45)>, <Segment(277.59, 278.89)>, <Segment(290.06, 291.15)>, <Segment(317.63, 318.04)>, <Segment(322.23, 325.08)>, <Segment(329.04, 329.27)>, <Segment(330.22, 331.53)>, <Segment(331.58, 332.82)>, <Segment(347.99, 348.31)>, <Segment(357.32, 358.06)>, <Segment(358.58, 358.95)>, <Segment(359.99, 360.45)>, <Segment(377.48, 377.93)>, <Segment(403.81, 406.48)>, <Segment(420.14, 420.46)>, <Segment(428.95, 431.54)>, <Segment(451.06, 451.37)>, <Segment(453.94, 454.36)>, <Segment(459.86, 460.58)>, <Segment(461.46, 461.8)>, <Segment(465.15, 466.45)>, <Segment(469.29, 469.68)>, <Segment(481.37, 484.34)>, <Segment(486.72, 486.73)>, <Segment(488.48, 489.46)>, <Segment(489.97, 490.46)>, <Segment(492.57, 492.92)>, <Segment(493.31, 495.51)>, <Segment(506.18, 506.75)>, <Segment(513.11, 513.57)>, <Segment(517.36, 517.76)>, <Segment(517.8, 518.74)>, <Segment(521.06, 521.07)>, <Segment(524.22, 525.41)>, <Segment(527.83, 533.05)>, <Segment(537.03, 537.48)>, <Segment(538.09, 539.24)>, <Segment(539.97, 540.33)>, <Segment(544.12, 544.57)>, <Segment(555.68, 556.72)>, <Segment(558.16, 558.63)>, <Segment(566.14, 566.78)>, <Segment(572.01, 575.79)>, <Segment(576.74, 576.76)>, <Segment(577.04, 577.78)>, <Segment(578.58, 580.16)>, <Segment(580.39, 580.91)>, <Segment(585.68, 586.34)>, <Segment(587, 587.95)>, <Segment(590.86, 593.03)>, <Segment(605.68, 605.71)>, <Segment(613.3, 613.69)>, <Segment(617.96, 618.26)>, <Segment(625.55, 626.09)>, <Segment(627.79, 628)>, <Segment(628.29, 628.67)>, <Segment(634.31, 637.21)>, <Segment(639.09, 639.15)>, <Segment(641.81, 642.12)>, <Segment(652.91, 653.24)>, <Segment(653.7, 654.09)>, <Segment(658.19, 658.44)>, <Segment(664.08, 664.74)>, <Segment(665.4, 667.15)>, <Segment(667.7, 667.93)>, <Segment(668.66, 671.59)>, <Segment(678.08, 679.99)>, <Segment(682.07, 683.11)>, <Segment(691.65, 692.33)>, <Segment(693.82, 693.97)>, <Segment(702.97, 703.28)>, <Segment(706.82, 707.73)>, <Segment(710.26, 710.7)>, <Segment(714.21, 714.49)>, <Segment(717.25, 718.7)>, <Segment(720.41, 720.78)>, <Segment(741.37, 741.7)>, <Segment(746.33, 746.79)>, <Segment(749.44, 749.86)>, <Segment(777.8, 777.82)>, <Segment(782.9, 783.2)>, <Segment(784.26, 784.46)>, <Segment(790.95, 793.93)>, <Segment(796.52, 797.59)>, <Segment(801.89, 802.4)>, <Segment(803.39, 803.61)>, <Segment(804.08, 804.53)>, <Segment(804.79, 805.07)>, <Segment(805.49, 805.68)>])>"
]
},
"execution_count": 35,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"oracle_osd(test_file).get_timeline()"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Because the model was trained on 2s audio chunks and that test files are likely to be much longer than that, we wrap the `model` with an `Inference` instance: it will take care of sliding a 2s window over the whole file and aggregate the output of the model."
]
},
{
"cell_type": "code",
"execution_count": 36,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.audio import Inference\n",
"inference = Inference(pretrained_model)\n",
"osd_probability = inference(test_file)"
]
},
{
"cell_type": "code",
"execution_count": 37,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<pyannote.core.feature.SlidingWindowFeature at 0x7f659658d900>"
]
},
"execution_count": 37,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"osd_probability"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"## Pipeline"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Almost there! To obtain the final overlapped speech regions, we need to apply a detection threshold. \n",
"For that, we rely on the overlapped speech detection pipeline whose hyper-parameters are set manually:\n",
"- `onset=0.5`: mark region as `active` when probability goes above 0.5\n",
"- `offset=0.5`: switch back to `inactive` when probability goes below 0.5\n",
"- `min_duration_on=0.1`: remove `active` regions shorter than 100ms\n",
"- `min_duration_off=0.1`: fill `inactive` regions shorter than 100ms."
]
},
{
"cell_type": "code",
"execution_count": 38,
"metadata": {},
"outputs": [],
"source": [
"from pyannote.audio.pipelines import OverlappedSpeechDetection as OverlappedSpeechDetectionPipeline\n",
"pipeline = OverlappedSpeechDetectionPipeline(pretrained_model).instantiate(\n",
" {\"onset\": 0.5, \"offset\": 0.5, \"min_duration_on\": 0.1, \"min_duration_off\": 0.1})"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"Here we go:"
]
},
{
"cell_type": "code",
"execution_count": 39,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Timeline(uri=IS1009a, segments=[<Segment(57.2261, 57.7304)>, <Segment(61.9565, 62.4957)>, <Segment(76.0957, 76.4783)>, <Segment(81.0348, 81.4348)>, <Segment(159.522, 159.974)>, <Segment(221.052, 221.626)>, <Segment(225.939, 226.461)>, <Segment(252.774, 253)>, <Segment(278.67, 279.696)>, <Segment(290.096, 291.278)>, <Segment(322.148, 323.33)>, <Segment(324.13, 325.087)>, <Segment(330.2, 330.809)>, <Segment(360.148, 360.461)>, <Segment(377.609, 377.904)>, <Segment(396.061, 396.339)>, <Segment(404.183, 404.878)>, <Segment(419.783, 420.374)>, <Segment(428.913, 430.548)>, <Segment(453.939, 454.565)>, <Segment(465.661, 466.478)>, <Segment(488.374, 490.443)>, <Segment(494.53, 495.609)>, <Segment(502.722, 503.487)>, <Segment(503.748, 504.165)>, <Segment(513.07, 513.417)>, <Segment(517.296, 517.939)>, <Segment(524.2, 525.243)>, <Segment(527.8, 528.878)>, <Segment(530.739, 531.8)>, <Segment(537, 538.843)>, <Segment(544.113, 544.617)>, <Segment(557.226, 558.53)>, <Segment(566.078, 566.791)>, <Segment(571.957, 575.887)>, <Segment(576.722, 577.696)>, <Segment(578.53, 581.365)>, <Segment(587, 587.957)>, <Segment(590.53, 592.965)>, <Segment(625.557, 626.357)>, <Segment(627.8, 628.861)>, <Segment(634.809, 636.6)>, <Segment(639, 639.313)>, <Segment(652.861, 653.4)>, <Segment(658.148, 658.53)>, <Segment(664.217, 666.322)>, <Segment(667.696, 669.852)>, <Segment(670.357, 671.974)>, <Segment(678.096, 680.148)>, <Segment(682.217, 683.07)>, <Segment(691.609, 692.148)>, <Segment(693.713, 694.009)>, <Segment(714.113, 714.548)>, <Segment(720.374, 720.948)>, <Segment(730.913, 731.383)>, <Segment(741.296, 741.696)>, <Segment(746.322, 746.896)>, <Segment(749.383, 749.835)>, <Segment(782.739, 783.104)>, <Segment(784.235, 784.357)>, <Segment(791.487, 791.713)>, <Segment(792.548, 793.817)>, <Segment(803.748, 804.704)>])>"
]
},
"execution_count": 39,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"pipeline(test_file).get_timeline()"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"## Optimizing pipeline hyper-parameters"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"W can try to optimize the hyper-parameters (that we chose manually above) on the validation set to get better performance."
]
},
{
"cell_type": "code",
"execution_count": 40,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"{'loss': 0.7273619100616171,\n",
" 'params': {'onset': 0.4321866835452871,\n",
" 'offset': 0.33656592926710593,\n",
" 'min_duration_on': 0.11171631479481339,\n",
" 'min_duration_off': 0.07311754878841395}}"
]
},
"execution_count": 40,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from pyannote.pipeline import Optimizer\n",
"validation_files = list(protocol.development())\n",
"optimizer = Optimizer(pipeline)\n",
"optimizer.tune(validation_files, n_iterations=200, show_progress=False)"
]
},
{
"attachments": {},
"cell_type": "markdown",
"metadata": {},
"source": [
"There you go: better hyper-parameters that should lead to better results!"
]
},
{
"cell_type": "code",
"execution_count": 41,
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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",
"text/plain": [
"<Timeline(uri=IS1009a, segments=[<Segment(56.913, 57.1217)>, <Segment(57.2087, 57.7826)>, <Segment(61.9391, 62.5478)>, <Segment(76.0957, 76.5304)>, <Segment(81.0174, 81.4696)>, <Segment(87.5043, 87.8522)>, <Segment(145.574, 145.8)>, <Segment(159.504, 160.061)>, <Segment(220.791, 221.887)>, <Segment(225.939, 226.513)>, <Segment(252.739, 253.087)>, <Segment(278.652, 279.8)>, <Segment(290.096, 291.4)>, <Segment(322.13, 323.574)>, <Segment(324.13, 325.348)>, <Segment(330.183, 330.93)>, <Segment(332.339, 332.513)>, <Segment(360.043, 360.6)>, <Segment(377.591, 377.974)>, <Segment(386.774, 387)>, <Segment(395.957, 396.409)>, <Segment(404.113, 404.913)>, <Segment(419.765, 420.426)>, <Segment(428.896, 430.774)>, <Segment(453.922, 454.67)>, <Segment(465.661, 466.565)>, <Segment(488.357, 490.548)>, <Segment(494.426, 495.696)>, <Segment(502.67, 504.339)>, <Segment(513.07, 513.504)>, <Segment(517.296, 517.991)>, <Segment(524.183, 525.383)>, <Segment(527.8, 528.93)>, <Segment(530.704, 532.043)>, <Segment(536.983, 539.035)>, <Segment(544.113, 544.687)>, <Segment(557.209, 558.617)>, <Segment(566.078, 566.861)>, <Segment(571.939, 576.2)>, <Segment(576.722, 577.835)>, <Segment(578.496, 581.452)>, <Segment(586.983, 588.009)>, <Segment(590.496, 593.104)>, <Segment(625.557, 626.409)>, <Segment(627.8, 628.913)>, <Segment(634.27, 636.774)>, <Segment(638.965, 639.417)>, <Segment(652.861, 653.522)>, <Segment(658.13, 658.6)>, <Segment(664.165, 666.53)>, <Segment(667.696, 672.078)>, <Segment(678.078, 680.2)>, <Segment(682.217, 683.226)>, <Segment(691.591, 692.235)>, <Segment(693.713, 694.096)>, <Segment(706.339, 706.722)>, <Segment(706.809, 707.035)>, <Segment(714.113, 714.6)>, <Segment(720.374, 721.035)>, <Segment(730.878, 731.47)>, <Segment(741.296, 741.73)>, <Segment(746.322, 746.948)>, <Segment(749.383, 749.922)>, <Segment(751.643, 751.852)>, <Segment(782.687, 783.209)>, <Segment(784.2, 784.443)>, <Segment(791.452, 792.443)>, <Segment(792.548, 793.939)>, <Segment(803.452, 805.174)>])>"
]
},
"execution_count": 41,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"optimized_pipeline = OverlappedSpeechDetectionPipeline(pretrained_model).instantiate(optimizer.best_params)\n",
"optimized_pipeline(test_file).get_timeline()"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.13"
}
},
"nbformat": 4,
"nbformat_minor": 4
}