add files

app.py

@@ -0,0 +1,189 @@
+import json
+import gradio as gr
+import matplotlib.pyplot as plt
+import numpy as np
+import os
+import requests
+from config import Config
+from model import BirdAST
+import torch
+import librosa
+import noisereduce as nr
+import pandas as pd
+import torch.nn.functional as F
+import random
+from torchaudio.compliance import kaldi
+from torchaudio.functional import resample
+from transformers import ASTFeatureExtractor
+
+#TAG = "gaunernst/vit_base_patch16_1024_128.audiomae_as2m_ft_as20k"
+#MODEL = timm.create_model(f"hf_hub:{TAG}", pretrained=True).eval()
+
+#LABEL_URL = "https://huggingface.co/datasets/huggingface/label-files/raw/main/audioset-id2label.json"
+#AUDIOSET_LABELS = list(json.loads(requests.get(LABEL_URL).content).values())
+
+FEATURE_EXTRACTOR  = ASTFeatureExtractor()
+
+def plot_mel(sr, x):
+    mel_spec = librosa.feature.melspectrogram(y=x, sr=sr, n_mels=224, fmax=10000)
+    mel_spec_db = librosa.power_to_db(mel_spec, ref=np.max)
+    mel_spec_db = (mel_spec_db - mel_spec_db.min()) / (mel_spec_db.max() - mel_spec_db.min()) # normalize spectrogram to [0,1]
+    mel_spec_db = np.stack([mel_spec_db, mel_spec_db, mel_spec_db], axis=-1)  # Convert to 3-channel
+    fig, ax = plt.subplots(nrows=1, ncols=1, sharex=True)
+    librosa.display.specshow(mel_spec_db[:, :, 0], sr=sr, x_axis='time', y_axis='mel', fmin = 0, fmax=10000, ax = ax)
+    return fig
+
+def plot_wave(sr, x):
+    ry = nr.reduce_noise(y=x, sr=sr)
+    fig, ax = plt.subplots(2, 1, figsize=(12, 8))
+
+    # Plot the original waveform
+    librosa.display.waveshow(x, sr=sr, ax=ax[0])
+    ax[0].set(title='Original Waveform')
+    ax[0].set_xlabel('Time (s)')
+    ax[0].set_ylabel('Amplitude')
+
+    # Plot the noise-reduced waveform
+    librosa.display.waveshow(ry, sr=sr, ax=ax[1])
+    ax[1].set(title='Noise Reduced Waveform')
+    ax[1].set_xlabel('Time (s)')
+    ax[1].set_ylabel('Amplitude')
+
+    plt.tight_layout()
+    return fig
+
+def predict(audio, start, end):
+    sr, x = audio
+    x = np.array(x, dtype=np.float64)
+    res = preprocess_for_inference(x, sr)
+    
+    if start >= end:
+        raise gr.Error(f"`start` ({start}) must be smaller than end ({end}s)")
+    
+    if x.shape[0] < start * sr:
+        raise gr.Error(f"`start` ({start}) must be smaller than audio duration ({x.shape[0] / sr:.0f}s)")
+    
+    if x.shape[0] > end * sr:
+        end = x.shape[0]/(1.0*sr)
+    
+    fig1 = plot_mel(sr, x)
+    fig2 = plot_wave(sr, x)
+
+
+    return res, res, fig1, fig2
+
+def download_model(url, model_path):
+    if not os.path.exists(model_path):
+        response = requests.get(url)
+        response.raise_for_status()  # Ensure the request was successful
+        with open(model_path, 'wb') as f:
+            f.write(response.content)
+
+# Model URL and path
+model_url = 'https://huggingface.co/shiyi-li/BirdAST/resolve/main/BirdAST_Baseline_fold_1.pth'
+model_path = 'BirdAST_Baseline_fold_1.pth'
+download_model(model_url, model_path)
+
+# Load the model (assumes you have the model architecture defined)
+eval_model = BirdAST(Config().backbone_name, Config().n_classes, n_mlp_layers=1, activation='silu')
+state_dict = torch.load('BirdAST_Baseline_fold_1.pth', map_location='cpu')
+eval_model.load_state_dict(state_dict)
+
+# Set to evaluation mode
+eval_model.eval()
+
+# Load the species mapping
+label_mapping = pd.read_csv('label_mapping.csv')
+species_id_to_name = {row['species_id']: row['scientific_name'] for index, row in label_mapping.iterrows()}
+
+
+def preprocess_for_inference(audio_arr, sr):
+    spec = FEATURE_EXTRACTOR(audio_arr, sampling_rate=sr, padding="max_length", return_tensors="pt")
+    input_values = spec['input_values'] #.squeeze(0) 
+    #print(input)
+    #print(input.shape)
+    results = []
+    with torch.no_grad():
+        # Perform model evaluation
+        output = eval_model(input_values)
+        predict_score = F.softmax(output['logits'], dim=1)
+        
+        # Get the top 10 predictions
+        topk_values, topk_indices = torch.topk(predict_score, 10, dim=1)
+        
+        # Map indices to species names and probabilities
+        for idx, scores in zip(topk_indices[0], topk_values[0]):
+            species_name = species_id_to_name[idx.item()]
+            probability = scores.item()
+            results.append([species_name, probability])
+            
+    return results
+
+DESCRIPTION = """
+Bird audio classification using SOTA Voice of Jungle Technology.
+"""
+
+"""
+with gr.Blocks() as demo:
+    submit_btn = gr.Button("Submit")
+demo = gr.Interface(
+    title="Bird audio classification",
+    description=DESCRIPTION,
+    fn=predict,
+    inputs=["audio", "number", "number"],
+    outputs=[
+        gr.Dataframe(headers=["class", "score"], row_count=10, label="prediction"),
+        gr.Plot(label="waveform"),
+        gr.Plot(label="spectrogram"),
+    ],
+    examples=[
+        ["312_Cissopis_leverinia_1.wav", 0, 5],
+        ["1094_Pionus_fuscus_2.wav", 0, 10],
+    ],
+)
+"""
+
+
+css = """
+.number-input {
+    height: 100%;
+    padding-bottom: 60px; /* Adust the value as needed for more or less space */
+}
+.full-height {
+    height: 100%;
+}
+.column-container {
+    height: 100%;
+}
+"""
+with gr.Blocks(css = css) as demo:
+    gr.Markdown("# Bird audio classification")
+    gr.Markdown(DESCRIPTION)
+    
+    with gr.Row():
+        with gr.Column(elem_classes="column-container"):
+            start_time_input = gr.Number(label="Start Time", value=0, elem_classes="number-input full-height")
+            end_time_input = gr.Number(label="End Time", value=1, elem_classes="number-input full-height")
+        with gr.Column():
+            audio_input = gr.Audio(label="Input Audio", elem_classes="full-height")
+  
+
+    with gr.Row():
+        raw_class_output = gr.Dataframe(headers=["class", "score"], row_count=10, label="Class Prediction")
+        species_output = gr.Dataframe(headers=["class", "score"], row_count=10, label="Species Prediction")
+        
+    with gr.Row():
+        waveform_output = gr.Plot(label="Waveform")
+        spectrogram_output = gr.Plot(label="Spectrogram")
+    
+    gr.Examples(
+        examples=[
+            ["312_Cissopis_leverinia_1.wav", 0, 5],
+            ["1094_Pionus_fuscus_2.wav", 0, 10],
+        ],
+        inputs=[audio_input, start_time_input, end_time_input]
+    )
+    
+    gr.Button("Predict").click(predict, [audio_input, start_time_input, end_time_input], [raw_class_output, species_output, waveform_output, spectrogram_output])
+
+demo.launch(share = True)

config.py

@@ -0,0 +1,25 @@
+class Config:
+    model_name = "BirdAST_Baseline_5folds"
+    backbone_name = "MIT/ast-finetuned-audioset-10-10-0.4593"
+    n_classes = 728 # number of classes in the dataset
+    audio_sr = 16000 #Hz
+    segment_length = 10  #s
+    fft_window = 0.025 #s
+    hop_window_length = 0.01 #s
+    n_mels = 128
+    low_cut = 1000 #Hz
+    high_cut = 8000 #Hz
+    top_db = 100
+    batch_size = 4 
+    num_workers = 0
+    n_splits = 5
+    max_lr = 1e-5
+    epochs = 15
+    weight_decay = 0.01
+    lr_final_div = 1000
+    amp = True
+    grad_accum_steps = 1
+    max_grad_norm = 1e7
+    print_epoch_freq = 1
+    print_freq = 500
+    random_seed = 2046

label_mapping.csv

@@ -0,0 +1,729 @@
+scientific_name,species_id
+Accipiter_superciliosus,0
+Akletos_melanoceps,1
+Amazona_amazonica,2
+Amazona_diadema,3
+Amazona_farinosa,4
+Amazona_festiva,5
+Amblycercus_holosericeus,6
+Ammodramus_aurifrons,7
+Ammonastes_pelzelni,8
+Ampelion_rufaxilla,9
+Anabazenops_dorsalis,10
+Anairetes_parulus,11
+Andigena_hypoglauca,12
+Anhima_cornuta,13
+Anisognathus_lacrymosus,14
+Anthracothorax_viridigula,15
+Antrostomus_rufus,16
+Ara_ararauna,17
+Ara_chloropterus,18
+Ara_severus,19
+Aramides_cajaneus,20
+"Aramus_Vieillot,",21
+Aratinga_leucophthalma,22
+Ardea_cocoi,23
+Arremon_abeillei,24
+Arremon_assimilis,25
+Arremon_taciturnus,26
+Asthenes_flammulata,27
+Atlapetes_latinuchus,28
+Atticora_tibialis,29
+Attila_bolivianus,30
+Attila_cinnamomeus,31
+Attila_citriniventris,32
+Attila_spadiceus,33
+Automolus_infuscatus,34
+Automolus_ochrolaemus,35
+Automolus_paraensis,36
+Automolus_rufipileatus,37
+Automolus_subulatus,38
+Basileuterus_tristriatus,39
+Berlepschia_rikeri,40
+Boissonneaua_matthewsii,41
+Brotogeris_chrysoptera,42
+Brotogeris_cyanoptera,43
+Brotogeris_sanctithomae,44
+Brotogeris_versicolurus,45
+Bucco_capensis,46
+Buteo_nitidus,47
+Buteogallus_schistaceus,48
+Buteogallus_urubitinga,49
+Buthraupis_montana,50
+Cacicus_cela,51
+Cacicus_chrysonotus,52
+Cacicus_haemorrhous,53
+Cacicus_leucoramphus,54
+Calidris_fuscicollis,55
+Campephilus_melanoleucos,56
+Campephilus_pollens,57
+Campephilus_rubricollis,58
+Camptostoma_obsoletum,59
+Campylorhamphus_procurvoides,60
+Campylorhynchus_fasciatus,61
+Cantorchilus_leucotis,62
+Capito_auratus,63
+Capito_aurovirens,64
+Capito_niger,65
+Capsiempis_flaveola,66
+Carduelis_magellanica,67
+Caryothraustes_canadensis,68
+Catharus_fuscater,69
+Celeus_elegans,70
+Celeus_flavus,71
+Celeus_grammicus,72
+Celeus_torquatus,73
+Celeus_undatus,74
+Cephalopterus_ornatus,75
+Ceratopipra_erythrocephala,76
+Ceratotriccus_granadensis,77
+Cercomacra_cinerascens,78
+Cercomacra_manu,79
+Cercomacra_nigrescens,80
+Cercomacroides_fuscicauda,81
+Cercomacroides_serva,82
+Cercomacroides_tyrannina,83
+Certhiasomus_stictolaemus,84
+Certhiaxis_cinnamomeus,85
+Certhiaxis_mustelinus,86
+Chaetura_brachyura,87
+Chaetura_spinicaudus,88
+Chamaeza_mollissima,89
+Chamaeza_nobilis,90
+Charadrius_collaris,91
+Chelidoptera_tenebrosa,92
+Chionomesa_fimbriata,93
+Chiroxiphia_pareola,94
+Chlorestes_cyanus,95
+Chloroceryle_americana,96
+Chlorophanes_spiza,97
+Chlorophonia_pyrrhophrys,98
+Chlorornis_riefferii,99
+Chlorospingus_flavopectus,100
+Chlorostilbon_mellisugus,101
+Chlorostilbon_olivaresi,102
+Chordeiles_pusillus,103
+Chrysomus_icterocephalus,104
+Ciccaba_albitarsis,105
+Cinnycerthia_peruana,106
+Cissopis_leverianus,107
+Cistothorus_platensis,108
+Cnemoscopus_rubrirostris,109
+Cnemotriccus_fuscatus,110
+Cnipodectes_subbrunneus,111
+Coccycua_minuta,112
+Coccyzus_melacoryphus,113
+Coeligena_violifer,114
+"Coereba_Vieillot,",115
+Colaptes_atricollis,116
+Colibri_coruscans,117
+Columba_speciosa,118
+Columbina_cruziana,119
+Columbina_minuta,120
+Conirostrum_bicolor,121
+Conirostrum_margaritae,122
+Conopias_parvus,123
+Conopias_trivirgatus,124
+Conopophaga_aurita,125
+Contopus_fumigatus,126
+Contopus_punensis,127
+Corythopis_torquatus,128
+Cranioleuca_antisiensis,129
+Cranioleuca_berlepschi,130
+Cranioleuca_gutturata,131
+Cranioleuca_muelleri,132
+Cranioleuca_vulpecula,133
+Cranioleuca_vulpina,134
+Crax_alector,135
+Crax_globulosa,136
+Crotophaga_ani,137
+Crotophaga_major,138
+Crypturellus_bartletti,139
+Crypturellus_casiquiare,140
+Crypturellus_cinereus,141
+Crypturellus_duidae,142
+Crypturellus_soui,143
+Crypturellus_undulatus,144
+Crypturellus_variegatus,145
+Cyanerpes_caeruleus,146
+Cyanicterus_cyanicterus,147
+Cyanocorax_heilprini,148
+Cyanocorax_violaceus,149
+Cyanocorax_yncas,150
+Cyanoloxia_rothschildii,151
+Cyanolyca_viridicyanus,152
+Cyclarhis_gujanensis,153
+Cymbilaimus_lineatus,154
+Cyphorhinus_arada,155
+Cyphorhinus_thoracicus,156
+Cyphos_macrodactylus,157
+Dacnis_albiventris,158
+Daptrius_ater,159
+Deconychura_longicauda,160
+Dendrexetastes_rufigula,161
+Dendrocincla_fuliginosa,162
+Dendrocincla_merula,163
+Dendrocolaptes_certhia,164
+Dendrocolaptes_hoffmannsi,165
+Dendrocolaptes_picumnus,166
+Dendrocygna_autumnalis,167
+Dendroplex_kienerii,168
+Deroptyus_accipitrinus,169
+Dichrozona_cincta,170
+Diglossa_cyanea,171
+Diglossa_mystacalis,172
+Diglossa_sittoides,173
+Donacobius_atricapilla,174
+Dromococcyx_phasianellus,175
+Drymophila_devillei,176
+Drymophila_striaticeps,177
+Dryocopus_lineatus,178
+Egretta_alba,179
+Egretta_thula,180
+Elaenia_flavogaster,181
+Elaenia_pallatangae,182
+Elaenia_pelzelni,183
+Elaenia_ruficeps,184
+Elliotomyia_chionogaster,185
+Empidonomus_varius,186
+Epinecrophylla_amazonica,187
+Epinecrophylla_gutturalis,188
+Epinecrophylla_haematonota,189
+Epinecrophylla_leucophthalma,190
+Epinecrophylla_ornata,191
+Eriocnemis_alinae,192
+Eubucco_richardsoni,193
+Euchrepomis_humeralis,194
+Eucometis_penicillata,195
+Euphonia_cayennensis,196
+Euphonia_chlorotica,197
+Euphonia_chrysopasta,198
+Euphonia_cyanocephala,199
+Euphonia_minuta,200
+Euphonia_plumbea,201
+Euphonia_rufiventris,202
+Euphonia_xanthogaster,203
+Eupsittula_pertinax,204
+Eurypygidae,205
+Falco_deiroleucus,206
+Falco_femoralis,207
+Falco_magnirostris,208
+Falco_rufigularis,209
+Florisuga_mellivora,210
+Formicarius_analis,211
+Formicarius_colma,212
+Formicivora_grisea,213
+Forpus_crassirostris,214
+Forpus_xanthops,215
+Frederickena_unduliger,216
+Frederickena_viridis,217
+Fringilla_capensis,218
+Fringilla_psaltria,219
+Furnarius_figulus,220
+Furnarius_leucopus,221
+Furnarius_minor,222
+Furnarius_torridus,223
+Galbacyrhynchus_leucotis,224
+Galbalcyrhynchus_purusianus,225
+Galbula_cyanescens,226
+Galbula_cyanicollis,227
+Galbula_dea,228
+Galbula_galbula,229
+Galbula_leucogastra,230
+Galbula_ruficauda,231
+Galbula_tombacea,232
+Gallinula_galeata,233
+Geothlypis_aequinoctialis,234
+Geranospiza_caerulescens,235
+Glaucidium_bolivianum,236
+Glaucidium_brasilianum,237
+Glaucidium_hardyi,238
+Glyphorynchus_spirurus,239
+Grallaria_blakei,240
+Grallaria_carrikeri,241
+Grallaria_obscura,242
+Grallaria_occabambae,243
+Grallaria_przewalskii,244
+Grallaria_quitensis,245
+Grallaria_ruficapilla,246
+Grallaria_squamigera,247
+Grallaricula_leymebambae,248
+Grallaricula_ochraceifrons,249
+Graydidascalus_brachyurus,250
+Gymnomystax_mexicanus,251
+Gymnopithys_leucaspis,252
+Gymnopithys_rufigula,253
+Gymnopithys_salvini,254
+Habia_rubica,255
+Hafferia_fortis,256
+Haplospiza_rustica,257
+Harpia_harpyja,258
+Heliangelus_amethysticollis,259
+Helicolestes_hamatus,260
+Heliornis_fulica,261
+Hemithraupis_flavicollis,262
+Hemitriccus_griseipectus,263
+Hemitriccus_inornatus,264
+Hemitriccus_josephinae,265
+Hemitriccus_minimus,266
+Hemitriccus_minor,267
+Hemitriccus_zosterops,268
+Henicorhina_leucophrys,269
+Henicorhina_leucoptera,270
+Henicorhina_leucosticta,271
+Herpetotheres_cachinnans,272
+Herpsilochmus_axillaris,273
+Herpsilochmus_dorsimaculatus,274
+Herpsilochmus_dugandi,275
+Herpsilochmus_parkeri,276
+Herpsilochmus_praedictus,277
+Herpsilochmus_stotzi,278
+Heterocercus_aurantiivertex,279
+Heterocercus_flavivertex,280
+Hoploxypterus_cayanus,281
+Hylexetastes_perrotii,282
+Hylexetastes_uniformis,283
+Hylocharis_sapphirina,284
+Hylopezus_macularius,285
+Hylophilus_brunneiceps,286
+Hylophilus_muscicapinus,287
+Hylophilus_olivaceus,288
+Hylophilus_pectoralis,289
+Hylophilus_semicinereus,290
+Hylophilus_thoracicus,291
+Hylophylax_naevius,292
+Hylophylax_punctulatus,293
+Hypnelus_bicinctus,294
+Hypocnemis_cantator,295
+Hypocnemis_flavescens,296
+Hypocnemis_hypoxantha,297
+Hypocnemis_peruviana,298
+Hypocnemis_rondoni,299
+Hypocnemis_striata,300
+Hypocnemoides_maculicauda,301
+Hypocnemoides_melanopogon,302
+Ibycter_americanus,303
+Icterus_cayanensis,304
+Ictinia_plumbea,305
+Incaspiza_laeta,306
+Incaspiza_watkinsi,307
+Inezia_subflava,308
+Iodopleura_isabellae,309
+Iridosornis_reinhardti,310
+Isleria_guttata,311
+Islerothraupis_luctuosa,312
+Jacamerops_aureus,313
+Jacana_jacana,314
+Kleinothraupis_atropileus,315
+Knipolegus_orenocensis,316
+Lampropsar_tanagrinus,317
+Lamprospiza_melanoleuca,318
+Lanio_fulvus,319
+Lanio_versicolor,320
+Laniocera_hypopyrra,321
+Larus_serranus,322
+Laterallus_exilis,323
+Lathrotriccus_euleri,324
+Leistes_bellicosus,325
+Lepidocolaptes_duidae,326
+Lepidocolaptes_fatimalimae,327
+Lepidocolaptes_fuscicapillus,328
+Lepidothrix_coronata,329
+Lepidothrix_nattereri,330
+Lepidothrix_serena,331
+Leptodon_cayanensis,332
+Leptopogon_taczanowskii,333
+Leptotila_rufaxilla,334
+Leptotila_verreauxi,335
+Leuconotopicus_fumigatus,336
+Leucopternis_kuhli,337
+Liosceles_thoracicus,338
+Lipaugus_vociferans,339
+Loddigesia_mirabilis,340
+Lophostrix_cristata,341
+Lophotriccus_galeatus,342
+Lophotriccus_vitiosus,343
+Lurocalis_rufiventris,344
+Malacoptila_fusca,345
+Manacus_manacus,346
+Margarornis_squamiger,347
+Mazaria_propinqua,348
+Mecocerculus_stictopterus,349
+Megascops_albogularis,350
+Megascops_choliba,351
+Megascops_ingens,352
+Megascops_petersoni,353
+Megascops_watsonii,354
+Megastictus_margaritatus,355
+Melanerpes_cruentatus,356
+Melanopareia_maranonica,357
+Micrastur_gilvicollis,358
+Micrastur_mintoni,359
+Micrastur_mirandollei,360
+Micrastur_ruficollis,361
+Micrastur_semitorquatus,362
+Microbates_collaris,363
+Microcerculus_bambla,364
+Microcerculus_marginatus,365
+Microrhopias_quixensis,366
+Milvago_chimachima,367
+Mimus_longicaudatus,368
+Mionectes_oleagineus,369
+Mionectes_striaticollis,370
+Mitrephanes_olivaceus,371
+Monasa_atra,372
+Monasa_morphoeus,373
+Monasa_nigrifrons,374
+Muscicapa_similis,375
+Myadestes_ralloides,376
+Myiarchus_cephalotes,377
+Myiarchus_swainsoni,378
+Myiarchus_tuberculifer,379
+Myiarchus_tyrannulus,380
+Myioborus_melanocephalus,381
+Myioborus_miniatus,382
+Myiodynastes_maculatus,383
+Myiopagis_caniceps,384
+Myiopagis_flavivertex,385
+Myiopagis_gaimardii,386
+Myiophobus_cryptoxanthus,387
+Myiornis_ecaudatus,388
+Myiotheretes_fumigatus,389
+Myiothlypis_coronata,390
+Myiothlypis_luteoviridis,391
+Myiothlypis_nigrocristata,392
+Myiozetetes_cayanensis,393
+Myiozetetes_luteiventris,394
+Myornis_senilis,395
+Myrmeciza_atrothorax,396
+Myrmeciza_longipes,397
+Myrmelastes_humaythae,398
+Myrmelastes_hyperythrus,399
+Myrmelastes_leucostigma,400
+Myrmelastes_rufifacies,401
+Myrmoborus_leucophrys,402
+Myrmoborus_lugubris,403
+Myrmoborus_melanurus,404
+Myrmoborus_myotherinus,405
+Myrmochanes_hemileucus,406
+Myrmoderus_ferruginea,407
+Myrmornis_torquata,408
+Myrmothera_campanisona,409
+Myrmotherula_ambigua,410
+Myrmotherula_assimilis,411
+Myrmotherula_axillaris,412
+Myrmotherula_brachyura,413
+Myrmotherula_cherriei,414
+Myrmotherula_iheringi,415
+Myrmotherula_klagesi,416
+Myrmotherula_longipennis,417
+Myrmotherula_menetriesii,418
+Nasica_longirostris,419
+Nemosia_pileata,420
+Neoctantes_niger,421
+Neopelma_chrysocephalum,422
+Neopipo_cinnamomea,423
+Nesotriccus_tumbezanus,424
+Nonnula_sclateri,425
+Notharchus_hyperrhynchus,426
+Notharchus_ordii,427
+Notharchus_tectus,428
+Nothocrax_urumutum,429
+Nyctalops_stygius,430
+Nyctibius_grandis,431
+Nyctibius_leucopterus,432
+Nyctidromus_albicollis,433
+Nyctiprogne_leucopyga,434
+Nystactes_tamatia,435
+Nystalus_obamai,436
+Ochthoeca_cinnamomeiventris,437
+Ochthoeca_diadema,438
+Ochthoeca_pulchella,439
+Ochthoeca_thoracica,440
+Ocreatus_underwoodii,441
+Odontophorus_gujanensis,442
+Onychorhynchus_coronatus,443
+Opisthocomidae,444
+Ornithion_inerme,445
+Ortalis_guttata,446
+Ortalis_motmot,447
+Pachyramphus_castaneus,448
+Pachyramphus_marginatus,449
+Pachyramphus_minor,450
+Pachyramphus_polychopterus,451
+Pachyramphus_rufus,452
+Pachyramphus_surinamus,453
+Pachyramphus_versicolor,454
+Pachysylvia_hypoxantha,455
+Pardirallus_sanguinolentus,456
+Paroaria_gularis,457
+Patagioenas_cayennensis,458
+Patagioenas_plumbea,459
+Patagioenas_subvinacea,460
+Penelope_jacquacu,461
+Penelope_montagnii,462
+Percnostola_rufifrons,463
+Perissocephalus_tricolor,464
+Phacellodomus_rufifrons,465
+Phaeomyias_murina,466
+Phaeothlypis_fulvicauda,467
+Phaeothlypis_rivularis,468
+Phaethornis_guy,469
+Phaethornis_hispidus,470
+Phaethornis_malaris,471
+Phaethornis_ruber,472
+Phaethornis_rupurumii,473
+Phaethornis_superciliosus,474
+Phaetusa_simplex,475
+Pharomachrus_auriceps,476
+Pharomachrus_pavoninus,477
+Pheucticus_chrysogaster,478
+Pheugopedius_coraya,479
+Pheugopedius_euophrys,480
+Pheugopedius_genibarbis,481
+Pheugopedius_paucimaculatus,482
+Philydor_erythropterum,483
+Philydor_pyrrhodes,484
+Phlegopsis_erythroptera,485
+Phlegopsis_nigromaculata,486
+Phoenicircus_carnifex,487
+Phoenicircus_nigricollis,488
+Phyllaemulor_bracteatus,489
+Phyllomyias_plumbeiceps,490
+Phylloscartes_gualaquizae,491
+Phylloscartes_ventralis,492
+Piaya_cayana,493
+Piaya_melanogaster,494
+Piculus_chrysochloros,495
+Piculus_flavigula,496
+Picumnus_castelnau,497
+Picumnus_exilis,498
+Picumnus_lafresnayi,499
+Picumnus_steindachneri,500
+Picumnus_varzeae,501
+Pionites_melanocephalus,502
+Pionus_fuscus,503
+Pionus_menstruus,504
+Pipile_cumanensis,505
+Pipra_aureola,506
+Pipra_pipra,507
+Pipraeidea_bonariensis,508
+Pipreola_arcuata,509
+Piprites_chloris,510
+Piranga_hepatica,511
+Pitangus_lictor,512
+Pitangus_sulphuratus,513
+Pithys_albifrons,514
+Platyrinchus_coronatus,515
+Platyrinchus_platyrhynchos,516
+Poecilotriccus_calopterus,517
+Poecilotriccus_capitalis,518
+Poecilotriccus_latirostris,519
+Poecilotriccus_luluae,520
+Poecilotriccus_senex,521
+Polioptila_attenboroughi,522
+Polioptila_guianensis,523
+Polioptila_plumbea,524
+Polytmus_theresiae,525
+Progne_subis,526
+Psarocolius_angustifrons,527
+Psarocolius_bifasciatus,528
+Psarocolius_decumanus,529
+Psarocolius_viridis,530
+Pseudastur_albicollis,531
+Pseudocolaptes_boissonneauii,532
+Pseudotriccus_ruficeps,533
+Psittacara_frontatus,534
+Psittacara_mitratus,535
+Psophia_crepitans,536
+Pteroglossus_aracari,537
+Pteroglossus_azara,538
+Pteroglossus_beauharnaisii,539
+Pteroglossus_castanotis,540
+Pulsatrix_perspicillata,541
+Pygiptila_stellaris,542
+Pygochelidon_melanoleuca,543
+Pyriglena_maura,544
+Pyrilia_aurantiocephala,545
+Pyrilia_barrabandi,546
+Pyrilia_caica,547
+Pyrrhomyias_cinnamomeus,548
+Pyrrhura_roseifrons,549
+Querula_purpurata,550
+Ramphastos_sulfuratus,551
+Ramphastos_tucanus,552
+Ramphastos_vitellinus,553
+Ramphocaenus_melanurus,554
+Ramphocelus_carbo,555
+Ramphocelus_nigrogularis,556
+Ramphotrigon_megacephalum,557
+Ramphotrigon_ruficauda,558
+Rhegmatorhina_berlepschi,559
+Rhegmatorhina_cristata,560
+Rhegmatorhina_hoffmannsi,561
+Rhegmatorhina_melanosticta,562
+Rhynchocyclus_olivaceus,563
+Rhytipterna_immunda,564
+Rhytipterna_simplex,565
+Rostrhamus_sociabilis,566
+Rufirallus_castaneiceps,567
+Rufirallus_viridis,568
+Rupicola_rupicola,569
+Rynchops_niger,570
+Sakesphorus_canadensis,571
+Sakesphorus_luctuosus,572
+Saltator_coerulescens,573
+Saltator_grossus,574
+Saltator_maximus,575
+Saltator_striatipectus,576
+Schiffornis_major,577
+Schiffornis_turdina,578
+Schistocichla_schistacea,579
+Schizoeaca_fuliginosa,580
+Sciaphylax_castanea,581
+Sciaphylax_hemimelaena,582
+Sclateria_naevia,583
+Sclerurus_obscurior,584
+Sclerurus_rufigularis,585
+Scytalopus_altirostris,586
+Scytalopus_atratus,587
+Scytalopus_femoralis,588
+Scytalopus_intermedius,589
+Scytalopus_parvirostris,590
+Selenidera_gouldii,591
+Selenidera_nattereri,592
+Selenidera_piperivora,593
+Selenidera_reinwardtii,594
+Sericossypha_albocristata,595
+Serpophaga_hypoleuca,596
+Serpophaga_subcristata,597
+Sicalis_columbiana,598
+Sirystes_sibilator,599
+Sirystes_subcanescens,600
+Sittasomus_griseicapillus,601
+Spinus_olivaceus,602
+Spizaetus_ornatus,603
+Sporophila_angolensis,604
+Sporophila_bouvronides,605
+Sporophila_castaneiventris,606
+Sporophila_fringilloides,607
+Sporophila_luctuosa,608
+Sporophila_simplex,609
+Sternula_superciliaris,610
+Stilpnia_viridicollis,611
+Sturnella_militaris,612
+Sublegatus_obscurior,613
+Synallaxis_albescens,614
+Synallaxis_albigularis,615
+Synallaxis_azarae,616
+Synallaxis_gujanensis,617
+Synallaxis_hypospodia,618
+Synallaxis_maranonica,619
+Synallaxis_rutilans,620
+Synallaxis_unirufa,621
+Syndactyla_rufosuperciliata,622
+Syndactyla_ucayalae,623
+Tachycineta_albiventer,624
+Tachyphonus_cristatus,625
+Tachyphonus_phoenicius,626
+Tachyphonus_surinamus,627
+Talaphorus_chlorocercus,628
+Tangara_cayana,629
+Tangara_chilensis,630
+Tangara_nigroviridis,631
+Tangara_punctata,632
+Tangara_varia,633
+Tangara_velia,634
+Tapera_naevia,635
+Taraba_major,636
+Terenura_spodioptila,637
+Tersina_viridis,638
+Thamnomanes_ardesiacus,639
+Thamnomanes_caesius,640
+Thamnomanes_saturninus,641
+Thamnomanes_schistogynus,642
+Thamnophilus_aethiops,643
+Thamnophilus_amazonicus,644
+Thamnophilus_bernardi,645
+Thamnophilus_cryptoleucus,646
+Thamnophilus_doliatus,647
+Thamnophilus_murinus,648
+Thamnophilus_nigrocinereus,649
+Thamnophilus_punctatus,650
+Thamnophilus_ruficapillus,651
+Thamnophilus_schistaceus,652
+Thamnophilus_stictocephalus,653
+Thamnophilus_unicolor,654
+Thaumasius_taczanowskii,655
+Thlypopsis_inornata,656
+Thlypopsis_superciliaris,657
+Thraupis_episcopus,658
+Thripadectes_scrutator,659
+Thripophaga_fusciceps,660
+Thryothorus_griseus,661
+Tiaris_obscurus,662
+Tinamus_major,663
+Tityra_cayana,664
+Tityra_semifasciata,665
+Todirostrum_chrysocrotaphum,666
+Todirostrum_maculatum,667
+Todirostrum_pictum,668
+Tolmomyias_assimilis,669
+Tolmomyias_flaviventris,670
+Tolmomyias_poliocephalus,671
+Tolmomyias_sulphurescens,672
+Tolmomyias_traylori,673
+Tolmomyias_viridiceps,674
+Topaza_pella,675
+Topaza_pyra,676
+Touit_purpuratus,677
+Troglodytes_aedon,678
+Troglodytes_solstitialis,679
+Trogon_collaris,680
+Trogon_curucui,681
+Trogon_melanurus,682
+Trogon_ramonianus,683
+Trogon_rufus,684
+Trogon_violaceus,685
+Trogon_viridis,686
+Tunchiornis_ochraceiceps,687
+Turdus_albicollis,688
+Turdus_hauxwelli,689
+Turdus_ignobilis,690
+Turdus_lawrencii,691
+Turdus_leucomelas,692
+Turdus_maranonicus,693
+Turdus_nigriceps,694
+Turdus_sanchezorum,695
+Turdus_serranus,696
+Tyranneutes_stolzmanni,697
+Tyranneutes_virescens,698
+Tyranniscus_cinereiceps,699
+Tyranniscus_uropygialis,700
+Tyrannulus_elatus,701
+Tyrannus_melancholicus,702
+Tyrannus_savana,703
+Uranomitra_franciae,704
+Uromyias_agraphia,705
+Uropsalis_segmentata,706
+Veniliornis_cassini,707
+Vireo_chivi,708
+Vireolanius_leucotis,709
+Volatinia_jacarina,710
+Wetmorethraupis_sterrhopteron,711
+Willisornis_poecilinotus,712
+Xenoglaux_loweryi,713
+Xenopipo_atronitens,714
+Xenops_milleri,715
+Xenops_minutus,716
+Xiphocolaptes_promeropirhynchus,717
+Xipholena_punicea,718
+Xiphorhynchus_elegans,719
+Xiphorhynchus_guttatus,720
+Xiphorhynchus_obsoletus,721
+Xiphorhynchus_ocellatus,722
+Xiphorhynchus_pardalotus,723
+Xiphorhynchus_picus,724
+Zimmerius_acer,725
+Zimmerius_gracilipes,726
+Zimmerius_viridiflavus,727

model.py

@@ -0,0 +1,39 @@
+import torch
+from torch.utils.data import Dataset, DataLoader
+from torch import nn
+from transformers import ASTConfig, ASTFeatureExtractor, ASTModel
+
+class BirdAST(nn.Module):
+    
+    def __init__(self, backbone_name, n_classes, n_mlp_layers=1, activation='silu'):
+        super(BirdAST, self).__init__()
+        
+        # pre-trained backbone
+        backbone_config = ASTConfig.from_pretrained(backbone_name)
+        self.ast = ASTModel.from_pretrained(backbone_name, config=backbone_config)
+        self.hidden_size = backbone_config.hidden_size
+        
+        # set activation functions
+        if activation == 'relu':
+            self.activation = nn.ReLU()
+        elif activation == 'silu':
+            self.activation = nn.SiLU()
+        else:
+            raise ValueError("Unsupported activation function. Choose 'relu' or 'silu'.")
+        
+        # define MLP layers with activation
+        layers = []
+        for _ in range(n_mlp_layers):
+            layers.append(nn.Linear(self.hidden_size, self.hidden_size))
+            layers.append(self.activation)
+        layers.append(nn.Linear(self.hidden_size, n_classes))
+        self.mlp = nn.Sequential(*layers)
+        
+    def forward(self, spectrogram):
+        # spectrogram: (batch_size, n_mels, n_frames)
+        # output: (batch_size, n_classes)
+        
+        ast_output = self.ast(spectrogram, output_hidden_states=False)
+        logits = self.mlp(ast_output.last_hidden_state[:, 0, :]) # Use the CLS token 
+        
+        return {'logits': logits}

requirements.txt

@@ -0,0 +1,11 @@
+gradio
+transformers
+matplotlib
+numpy
+requests
+timm
+pandas
+torch
+librosa
+noisereduce
+torchaudio