sync ms model

README.md

@@ -6,7 +6,7 @@ language:
 - en
 metrics:
 - accuracy
-pipeline_tag: image-classification
+pipeline_tag: audio-classification
 tags:
 - music
 - art
@@ -14,6 +14,12 @@ tags:
 
 The Classical and Ethnic Vocal Style Classification model aims to distinguish between classical and ethnic vocal styles, with all audio samples sung by professional vocalists. The model is fine-tuned using an audio dataset consisting of four categories, which has been pre-processed into spectrograms. Initially pretrained in the computer vision (CV) domain, the backbone network undergoes a fine-tuning process specifically designed for vocal style classification tasks. In this model, the pre-training on CV tasks provides a foundation for the network to learn general audio features, which are then adjusted during fine-tuning to adapt to the subtle differences between classical and ethnic vocal styles. The audio dataset, comprising samples from classical and various ethnic singing traditions, enables the model to capture unique patterns associated with each vocal style. Representing spectrograms as input allows the model to effectively analyze both the temporal and frequency components of the audio signals. Through the fine-tuning process, the model continuously enhances its ability to discriminate between sound representations and subtle stylistic differences between classical and ethnic styles. This specialized model holds significant potential in the music industry and cultural preservation, as it accurately categorizes vocal performances into these two broad categories. Its foundation in pre-trained computer vision principles demonstrates the versatility and adaptability of neural networks across different domains, enhancing the model's capability to capture complex features of vocal performances.
 
+## Usage
+```python
+from modelscope import snapshot_download
+model_dir = snapshot_download('ccmusic-database/bel_canto')
+```
+
 ## Maintenance
 ```bash
 GIT_LFS_SKIP_SMUDGE=1 git clone [email protected]:ccmusic-database/bel_canto
@@ -22,7 +28,6 @@ cd bel_canto
 
 ## Results
 A demo result of SqueezeNet fine-tuning:
-
 <style>
   #pianos td {
     vertical-align: middle !important;
@@ -35,20 +40,20 @@ A demo result of SqueezeNet fine-tuning:
 <table id="pianos">
     <tr>
         <th>Loss curve</th>
-        <td><img src="https://huggingface.co/ccmusic-database/bel_canto/resolve/main/squeezenet1_1_cqt_2024-01-30_00-57-26/loss.jpg"></td>
+        <td><img src="./loss.jpg"></td>
     </tr>
     <tr>
         <th>Training and validation accuracy</th>
-        <td><img src="https://huggingface.co/ccmusic-database/bel_canto/resolve/main/squeezenet1_1_cqt_2024-01-30_00-57-26/acc.jpg"></td>
+        <td><img src="./acc.jpg"></td>
     </tr>
     <tr>
         <th>Confusion matrix</th>
-        <td><img src="https://huggingface.co/ccmusic-database/bel_canto/resolve/main/squeezenet1_1_cqt_2024-01-30_00-57-26/mat.jpg"></td>
+        <td><img src="./mat.jpg"></td>
     </tr>
 </table>
 
 ## Mirror
-<https://www.modelscope.cn/models/ccmusic/bel_canto>
+<https://www.modelscope.cn/models/ccmusic-database/bel_canto>
 
 ## Reference
-<https://github.com/monet-joe/ccmusic_clstask_eval>
+[1] <https://github.com/monetjoe/ccmusic_eval>

alexnet_cqt_2024-02-20_06-54-06/acc.csv

@@ -1,41 +0,0 @@
-tra_acc_list,val_acc_list,lr_list
-69.27883363707367,70.20833333333333,0.001
-74.01718302525384,75.52083333333334,0.001
-77.4147357458995,78.33333333333333,0.001
-79.40640458214007,81.25,0.001
-77.15438687841709,78.33333333333333,0.001
-80.60400937255923,79.79166666666667,0.001
-82.27024212444675,82.60416666666667,0.001
-82.46550377505858,80.72916666666666,0.001
-83.72819578234835,82.39583333333333,0.001
-86.0192658161937,85.3125,0.0001
-86.27961468367613,83.64583333333333,0.0001
-86.00624837281958,85.3125,0.0001
-85.6547774017183,84.89583333333334,0.0001
-86.95652173913044,85.625,0.0001
-86.696172871648,85.10416666666667,0.0001
-86.50091122103619,86.25,1e-05
-86.76126008851861,85.3125,1e-05
-86.76126008851861,84.47916666666667,1e-05
-86.82634730538922,86.45833333333334,1e-05
-86.93048685238219,85.625,1e-05
-86.2015100234314,84.27083333333333,1e-05
-86.72220775839625,85.41666666666666,1e-05
-87.03462639937516,84.6875,1e-05
-87.15178338974225,86.04166666666667,1e-05
-87.07367872949753,86.14583333333333,1e-05
-86.51392866441032,85.20833333333333,1.0000000000000002e-06
-86.8784170788857,85.72916666666667,1.0000000000000002e-06
-86.67013798489977,84.6875,1.0000000000000002e-06
-87.47721947409529,85.10416666666667,1.0000000000000002e-06
-86.51392866441032,85.625,1.0000000000000002e-06
-86.96953918250455,84.79166666666667,1.0000000000000002e-06
-87.3079927102317,85.72916666666667,1.0000000000000002e-07
-86.89143452225983,84.0625,1.0000000000000002e-07
-86.74824264514449,86.875,1.0000000000000002e-07
-86.78729497526686,86.66666666666667,1.0000000000000002e-07
-86.5659984379068,85.0,1.0000000000000002e-07
-86.68315542827388,85.10416666666667,1.0000000000000002e-07
-86.24056235355376,86.14583333333333,1.0000000000000004e-08
-86.3837542306691,84.79166666666667,1.0000000000000004e-08
-86.94350429575631,85.0,1.0000000000000004e-08

alexnet_cqt_2024-02-20_06-54-06/mat.csv

@@ -1,4 +0,0 @@
-1.321540062434963525e-01,0.000000000000000000e+00,2.705515088449531730e-02,0.000000000000000000e+00
-0.000000000000000000e+00,1.737773152965660650e-01,0.000000000000000000e+00,3.642039542143600539e-02
-2.705515088449531730e-02,0.000000000000000000e+00,1.654526534859521447e-01,4.162330905306972292e-03
-0.000000000000000000e+00,3.850156087408949240e-02,9.365244536940686357e-03,3.860561914672216433e-01

alexnet_cqt_2024-02-20_06-54-06/result.log

@@ -1,18 +0,0 @@
-              precision    recall  f1-score   support
-
-       m_bel      0.830     0.830     0.830       153
-       f_bel      0.819     0.827     0.823       202
-      m_folk      0.820     0.841     0.830       189
-      f_folk      0.905     0.890     0.897       417
-
-    accuracy                          0.857       961
-   macro avg      0.843     0.847     0.845       961
-weighted avg      0.858     0.857     0.858       961
-
-Backbone      : alexnet
-Spect type    : cqt
-Start time    : 2024-02-20 06:05:26
-Finish time   : 2024-02-20 06:53:56
-Time cost     : 2909s
-Full finetune : True
-Focal loss    : True

alexnet_cqt_2024-02-20_06-54-06/save.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5407c21bfb9fcbb5ccdd38efa2f276917e89d1007127e0ace598429566c01708
-size 58618479

alexnet_mel_2024-02-20_08-11-04/acc.csv

@@ -1,41 +0,0 @@
-tra_acc_list,val_acc_list,lr_list
-67.69070554543087,69.58333333333333,0.001
-73.48346784691486,74.6875,0.001
-73.71778182764905,75.41666666666667,0.001
-77.31059619890654,77.39583333333333,0.001
-79.19812548815412,78.95833333333333,0.001
-79.38036969539183,78.125,0.001
-81.71049205935954,80.72916666666666,0.001
-82.14006769070554,83.95833333333333,0.001
-83.91043998958605,81.97916666666667,0.001
-84.58734704504035,82.39583333333333,0.001
-84.26191096068733,85.0,0.001
-86.73522520177038,84.0625,0.001
-87.64644623795887,86.97916666666666,0.001
-88.16714397292372,87.08333333333333,0.0001
-87.86774277531893,84.47916666666667,0.0001
-88.08903931267899,85.83333333333333,0.0001
-88.7659463681333,87.91666666666667,0.0001
-89.39078365009112,85.72916666666667,0.0001
-89.23457432960167,87.1875,0.0001
-89.41681853683936,86.875,0.0001
-88.6487893777662,85.72916666666667,1e-05
-89.35173131996876,86.45833333333334,1e-05
-88.89612080187452,85.9375,1e-05
-89.32569643322051,86.77083333333333,1e-05
-89.23457432960167,87.08333333333333,1e-05
-89.61208018745118,85.20833333333333,1e-05
-89.29966154647228,86.25,1.0000000000000002e-06
-89.16948711273106,85.52083333333333,1.0000000000000002e-06
-88.79198125488155,85.83333333333333,1.0000000000000002e-06
-89.40380109346523,87.08333333333333,1.0000000000000002e-06
-89.15646966935694,87.8125,1.0000000000000002e-06
-89.1955219994793,86.875,1.0000000000000002e-06
-89.5600104139547,86.97916666666666,1.0000000000000002e-06
-89.10439989586045,85.9375,1.0000000000000002e-06
-89.22155688622755,85.52083333333333,1.0000000000000002e-06
-89.52095808383234,85.83333333333333,1.0000000000000002e-06
-89.31267898984639,85.83333333333333,1.0000000000000002e-06
-89.36474876334289,86.45833333333334,1.0000000000000002e-07
-88.85706847175214,85.41666666666666,1.0000000000000002e-07
-89.28664410309815,86.14583333333333,1.0000000000000002e-07

alexnet_mel_2024-02-20_08-11-04/mat.csv

@@ -1,4 +0,0 @@
-1.383975026014568066e-01,0.000000000000000000e+00,2.081165452653485973e-02,0.000000000000000000e+00
-0.000000000000000000e+00,1.758584807492195590e-01,0.000000000000000000e+00,3.433922996878251838e-02
-2.185223725286160323e-02,0.000000000000000000e+00,1.696149843912591049e-01,5.202913631633714932e-03
-0.000000000000000000e+00,3.433922996878251838e-02,7.284079084287201078e-03,3.922996878251820974e-01

alexnet_mel_2024-02-20_08-11-04/result.log

@@ -1,18 +0,0 @@
-              precision    recall  f1-score   support
-
-       m_bel      0.864     0.869     0.866       153
-       f_bel      0.837     0.837     0.837       202
-      m_folk      0.858     0.862     0.860       189
-      f_folk      0.908     0.904     0.906       417
-
-    accuracy                          0.876       961
-   macro avg      0.867     0.868     0.867       961
-weighted avg      0.876     0.876     0.876       961
-
-Backbone      : alexnet
-Spect type    : mel
-Start time    : 2024-02-20 07:13:05
-Finish time   : 2024-02-20 08:10:52
-Time cost     : 3466s
-Full finetune : True
-Focal loss    : True

alexnet_mel_2024-02-20_08-11-04/save.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:cf9da8e4eea04186a9d92a9d960cfb2bb3948350ce50de4aa4704d46f0a99962
-size 58618479

convnext_tiny_cqt_4cls_2024-02-25_21-30-35/acc.csv

@@ -1,41 +0,0 @@
-tra_acc_list,val_acc_list,lr_list
-51.44493621452747,53.22916666666667,0.001
-58.05779744858109,61.5625,0.001
-61.76776881020568,64.16666666666667,0.001
-63.91564696693569,64.89583333333333,0.001
-68.00312418640979,71.25,0.001
-71.32257224681072,71.25,0.001
-73.66571205415256,75.52083333333334,0.001
-74.26451444936214,76.5625,0.001
-74.81124707107524,77.08333333333334,0.001
-74.03020046862795,75.10416666666667,0.001
-75.47513668315543,77.70833333333333,0.001
-76.39937516271804,76.77083333333333,0.001
-80.14839885446499,81.5625,0.001
-80.73418380630044,81.5625,0.001
-78.23483467846914,77.91666666666667,0.001
-81.21582921114293,81.5625,0.001
-83.59802134860713,83.125,0.001
-83.36370736787295,82.08333333333333,0.001
-82.71283519916688,82.08333333333333,0.001
-80.89039312678989,82.29166666666666,0.001
-83.71517833897423,82.8125,0.001
-83.3897422546212,82.91666666666667,0.001
-87.37307992710231,85.41666666666666,0.0001
-87.93282999218953,87.39583333333333,0.0001
-88.12809164280135,88.33333333333333,0.0001
-88.31033585003905,87.29166666666667,0.0001
-88.79198125488155,86.875,0.0001
-89.07836500911222,86.875,0.0001
-88.84405102837802,87.29166666666667,1e-05
-89.09138245248633,86.66666666666667,1e-05
-88.66180682114033,86.14583333333333,1e-05
-89.48190575370997,88.02083333333334,1e-05
-89.07836500911222,88.125,1e-05
-89.32569643322051,86.77083333333333,1e-05
-88.90913824524863,88.02083333333334,1.0000000000000002e-06
-88.71387659463682,86.97916666666666,1.0000000000000002e-06
-89.13043478260869,87.39583333333333,1.0000000000000002e-06
-89.05233012236397,88.125,1.0000000000000002e-06
-89.35173131996876,88.54166666666666,1.0000000000000002e-06
-88.97422546211924,87.1875,1.0000000000000002e-06

convnext_tiny_cqt_4cls_2024-02-25_21-30-35/mat.csv

@@ -1,4 +0,0 @@
-1.383975026014568066e-01,0.000000000000000000e+00,2.081165452653485973e-02,0.000000000000000000e+00
-0.000000000000000000e+00,1.789802289281997860e-01,0.000000000000000000e+00,3.121748178980228786e-02
-3.433922996878251838e-02,0.000000000000000000e+00,1.602497398543184237e-01,2.081165452653486146e-03
-0.000000000000000000e+00,2.809573361082206081e-02,6.243496357960457571e-03,3.995837669094692846e-01

convnext_tiny_cqt_4cls_2024-02-25_21-30-35/result.log

@@ -1,20 +0,0 @@
-              precision    recall  f1-score   support
-
-       m_bel      0.801     0.869     0.834       153
-       f_bel      0.864     0.851     0.858       202
-      m_folk      0.856     0.815     0.835       189
-      f_folk      0.923     0.921     0.922       417
-
-    accuracy                          0.877       961
-   macro avg      0.861     0.864     0.862       961
-weighted avg      0.878     0.877     0.877       961
-
-Class num     : 4
-Backbone      : convnext_tiny
-Pretrain src  : ImageNet1k_v1
-Data column   : cqt
-Start time    : 2024-02-25 18:37:32
-Finish time   : 2024-02-25 21:30:25
-Time cost     : 10373s
-Full finetune : True
-Focal loss    : True

convnext_tiny_cqt_4cls_2024-02-25_21-30-35/save.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:a3887ea8124ecb661e30d44ee24cfcf5666e8e85e8fa3bea2ad6449b00eea01a
-size 111985033

convnext_tiny_mel_4cls_2024-02-25_20-20-51/acc.csv

@@ -1,41 +0,0 @@
-tra_acc_list,val_acc_list,lr_list
-52.10882582660765,52.39583333333333,0.001
-54.68627961468368,52.81249999999999,0.001
-59.19031502212966,55.52083333333333,0.001
-62.50976308253059,60.10416666666667,0.001
-65.39963551158553,63.541666666666664,0.001
-68.45873470450402,66.5625,0.001
-71.02317104920594,70.625,0.001
-72.68940380109346,69.27083333333334,0.001
-72.24681072637334,69.89583333333333,0.001
-73.06690965894298,73.4375,0.001
-70.15100234313981,68.85416666666667,0.001
-77.27154386878418,75.3125,0.001
-78.29992189533975,75.72916666666667,0.001
-80.0312418640979,76.04166666666666,0.001
-80.94246290028639,77.91666666666667,0.001
-79.32829992189534,73.75,0.001
-83.5719864618589,79.89583333333333,0.0001
-83.7021608956001,78.85416666666667,0.0001
-84.41812028117678,81.97916666666667,0.0001
-85.23821921374642,82.29166666666666,0.0001
-84.977870346264,81.25,0.0001
-85.26425410049467,80.52083333333333,0.0001
-85.61572507159593,83.22916666666667,1e-05
-85.30330643061703,81.45833333333333,1e-05
-85.57667274147357,82.5,1e-05
-85.79796927883365,82.1875,1e-05
-85.53762041135121,79.6875,1e-05
-85.27727154386878,82.1875,1e-05
-85.30330643061703,83.125,1.0000000000000002e-06
-85.82400416558188,82.91666666666667,1.0000000000000002e-06
-85.82400416558188,81.35416666666667,1.0000000000000002e-06
-85.71986461858891,82.08333333333333,1.0000000000000002e-06
-85.49856808122884,82.1875,1.0000000000000002e-06
-85.02994011976048,83.4375,1.0000000000000002e-06
-85.26425410049467,82.39583333333333,1.0000000000000002e-07
-85.66779484509243,81.66666666666667,1.0000000000000002e-07
-86.43582400416558,82.08333333333333,1.0000000000000002e-07
-85.45951575110648,82.08333333333333,1.0000000000000002e-07
-85.51158552460298,81.875,1.0000000000000002e-07
-85.70684717521479,81.97916666666667,1.0000000000000002e-07

convnext_tiny_mel_4cls_2024-02-25_20-20-51/mat.csv

@@ -1,4 +0,0 @@
-1.196670135275754443e-01,0.000000000000000000e+00,2.705515088449531730e-02,0.000000000000000000e+00
-0.000000000000000000e+00,1.935483870967741882e-01,0.000000000000000000e+00,4.994797086368366057e-02
-4.266389177939645949e-02,0.000000000000000000e+00,1.602497398543184237e-01,5.202913631633714932e-03
-0.000000000000000000e+00,3.746097814776274543e-02,1.040582726326743073e-03,3.631633714880332930e-01

convnext_tiny_mel_4cls_2024-02-25_20-20-51/result.log

@@ -1,20 +0,0 @@
-              precision    recall  f1-score   support
-
-       m_bel      0.737     0.816     0.774       141
-       f_bel      0.838     0.795     0.816       234
-      m_folk      0.851     0.770     0.808       200
-      f_folk      0.868     0.904     0.886       386
-
-    accuracy                          0.837       961
-   macro avg      0.824     0.821     0.821       961
-weighted avg      0.838     0.837     0.836       961
-
-Class num     : 4
-Backbone      : convnext_tiny
-Pretrain src  : ImageNet1k_v1
-Data column   : mel
-Start time    : 2024-02-25 18:37:01
-Finish time   : 2024-02-25 20:20:45
-Time cost     : 6223s
-Full finetune : True
-Focal loss    : True

convnext_tiny_mel_4cls_2024-02-25_20-20-51/save.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:da52fc4b2dcdbbd4bd095962dd50110047246912118d30a3172311fe040aa951
-size 111985033

densenet201_cqt_4cls_2024-02-23_15-46-35.7z

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9af057a52f3935cac0d04e875e3493d319a07d53eb61c975cfdb458dfe066443
-size 69239073

densenet201_mel_4cls_2024-02-23_15-07-59/acc.csv

@@ -1,41 +0,0 @@
-tra_acc_list,val_acc_list,lr_list
-41.484375,41.354166666666664,0.001
-41.47135416666667,41.25,0.001
-41.510416666666664,41.66666666666667,0.001
-46.276041666666664,48.95833333333333,0.001
-49.075520833333336,48.854166666666664,0.001
-50.9375,50.31250000000001,0.001
-50.33854166666667,51.145833333333336,0.001
-51.70572916666667,50.625,0.001
-52.82552083333333,53.02083333333333,0.001
-53.76302083333333,52.5,0.001
-58.385416666666664,56.97916666666667,0.001
-57.942708333333336,56.97916666666667,0.001
-58.67187499999999,56.875,0.001
-59.23177083333333,59.27083333333333,0.001
-60.9375,59.166666666666664,0.001
-61.99218749999999,60.83333333333333,0.001
-63.697916666666664,63.33333333333333,0.001
-64.97395833333334,61.458333333333336,0.001
-66.30208333333333,64.27083333333333,0.001
-66.31510416666667,63.95833333333333,0.001
-67.99479166666667,64.6875,0.001
-66.18489583333333,66.875,0.001
-69.921875,66.45833333333333,0.001
-70.859375,70.10416666666667,0.001
-72.7734375,70.83333333333334,0.001
-71.11979166666667,68.125,0.001
-71.66666666666667,68.54166666666667,0.001
-73.4375,70.3125,0.001
-73.59375,72.1875,0.001
-74.53125,69.47916666666667,0.001
-74.44010416666667,72.08333333333333,0.001
-73.7890625,73.125,0.001
-75.44270833333333,73.85416666666667,0.001
-75.16927083333333,74.47916666666666,0.001
-76.43229166666666,75.3125,0.001
-76.00260416666667,73.125,0.001
-77.90364583333333,74.27083333333333,0.0001
-79.16666666666666,74.89583333333333,0.0001
-79.27083333333333,76.45833333333333,0.0001
-79.70052083333333,77.70833333333333,0.0001