I-Machine Learning for the Birds: Ukwakha I-Our Own Bird Vocalization Classifier

Ukucaciswa Iingcali usebenzisa iinkqubo ze-automated ukufundisa i-ecosystems ezininzi. Kwiimeko ze-forest and jungle areas, iinkcukacha ezisetyenziswa ukubonisa i-audio eyenziwa ukunceda ukucacisa iintlobo ezahlukeneyo ze-animals kunye ne-insects. Le nkcazelo ingasetyenziswa ukucaciswa okuphumeleleyo lokuquka iintlobo ngaphakathi kwimvelo efikelelekayo. Kwiimveliso ze-birds, i-Google Research ibonisa kwi-artikeli yayo ukuba "i-ecologists usebenzisa amafutha ukufumana iinkqubo yemveliso kunye neemveliso zempilo - umzekelo, ukuba kukho iimveliso ezininzi kwiimveliso, oku kuthetha ukuba kukho iimveliso ezininzi ezidlulileyo." Ukongezelela, zibonisa i-value ye-audio-based identification: "[Kuba] amafutha zihlanganisa kunye nokubonisa indawo ngeengoma kunye neengxaki, kunzima kakhulu ukucinga kubo nge-ear. autonomous recording units (ARUs) Ukuhlukanisa i-Birdsong in the Wild ngenxa ye-Classification Ukuhlukanisa i-Birdsong in the Wild ngenxa ye-Classification Kwangoku, i Iingcingo ezivela kwi Phantsi umnqweno we inkampani. I-ImageCLEF inikezela uphando kwi-anotation ye-cross-language kunye nokufumana iifoto kwiintlobo ezahlukeneyo. Umthamo we-competition yinto epheleleyo: yenza iimodeli ye-classification enokufuneka ngokunyaniseka iintlobo ze-bird ukusuka kwiphepha le-audio. I-BirdCLEF+ ye-2025 Kaggle Ukucinga I-BirdCLEF+ ye-2025 Ukucinga Okokuqala, umsebenzi kubonakala banxibelelana ngokufanelekileyo ngokufanelekileyo Kwakhona eyaziwa njenge . I-GBV i-classifier ifunyenwe kwi-11,000 iintlobo ze-bird yaye ngoko kubalulekile ukuba ifunyenwe njenge-model ye-classification. I-Google Bird Vocalization (GBV) I-Classifier ye-Google Perch I-Google Bird Vocalization (GBV) I-Classifier ye-Google Nangona kunjalo, i-competition ibandakanya iintlobo ze-bird ezikhoyo kwi-GBV classifier training set. Ngenxa yoko, i-GBV classifier ibandakanya kuphela i-~60% i-accuracy kwi-BirdCLEF+ 2025 competition test dataset. Ngenxa yoko kufuneka yenzelwe iimodeli ye-custom. Le nqakraza ibonisa indlela yokwakha i-classifier yakho ye-bird vocalization leyo ingasetyenziswa kunye ne-GBV classifier uklasifikisa i-selection ye-bird species. Le nqakraza isetyenziswa izixhobo zokuqala ezaziwayo kwi-Google Research I-Design isebenzisa i-BirdCLEF+ 2025 competition data set ekulungiseleni. iimveliso Ukuqeqesha Data Yintoni , kuquka iifayile ezihambelana, i-approximately 12 GB. Iifayile ezihambelanayo kunye neefayile ezihambelanayo iifayile ze-dataset ziquka: I-BirdCLEF+ 2025 i-training data set birdclef_2025 |__ train_audio |__ train_soundscapes |__ test_soundscapes recording_location.txt taxonomy.csv train.csv train_audio Yintoni I-catalogue yi-component eninzi ye-dataset, ebandakanya i-28,564 iingcebiso ze-audio ze-training audio format. Iingcebiso ze-audio zihlanganiswa kwi-sub-directories ezininzi zihlanganisa iintlobo zangaphakathi, njl: train_audio .ogg train_audio |__amakin1 |__ [AUDIO FILES] |__amekes |__ [AUDIO FILES] ... Yintoni ifayile ingasetyenziswa ukufumana iimpawu zenzululwazi kunye neengxaki ezivela kwiintlobo ze-bird ezidlulileyo kunye neengxaki ze-sub-directory, njl: taxonomy.csv SUB-DIRECTORY NAME SCIENTIFIC NAME COMMON NAME amakin1 Chloroceryle amazona Amazon Kingfisher amekes Falco sparverius American Kestrel ... I-dataset ye-competition ibandakanya i-206 iintlobo ze-animal, ngoko ke i-206 iingqungquthela. Iintlobo ze-63 ziyafumaneka Ukubandakanya i Iinkcukacha iindidi zokusetyenziswa ngokubanzi nge-numerical class identifier: Introduction Ngaba GBV Classifier Non-GBV 1139490, 1192948, 1194042, 126247, 1346504, 134933, 135045, 1462711, 1462737, 1564122, 21038, 21116, 21211, 22333, 22973, 22976, 24272, 24292, 24322, 41663, 41778, 41970, 42007, 42087, 42113, 46010, 47067, 476537, 476538, 48124, 50186, 517119, 523060, 528041, 52884, 548639, 555086, 555142, 566513, 64862, 65336, 65344, 65349, 65373, 65419, 65448, 65547, 65962, 66016, 66531, 66578, 66893, 67082, 67252, 714022, 715170, 787625, 81930, 868458, 963335, grasal4, verfly, y00678 Uninzi le Izifundo zihlanganiswa nge: Non-GBV Limited training data. Class , for example, only contains 2 audio recordings. By contrast, class , which is a “known” class, contains 89 recordings. 1139490 amakin1 GBV Poor recording quality. Highlighting class again, both training recordings are of poor quality with one being particularly difficult to discern. 1139490 Lezi zimo ezimbini zibonisa ubunzima phakathi kwezilwanyana ngokutsho ubungakanani umgangatho umgangatho umgangatho umgangatho umgangatho. Zonke iingcingo zokusebenza ze-audio kwiintlobo ezimbini iimveliso izifundo zihlanganisa nokufunda umntu, kwaye umdlali uthetha iingcingo kunye neengcingo efana ne-species of bird that was recorded and location of the recording. In most - - iimeko, iingcebiso zihlanganisa iingcebiso ze-bird ebhalisiweyo. GBV Non-GBV Kwaye akukho zonke Iingcingo ezisetyenziswa ukuyisombulula i-class imbalance kunye nokufumaneka kwe-anotations ye-human speech ziya kubhalwe kwi iinkcukacha Building the Classifier train_soundscapes Yintoni inkxaso ibandakanya malunga ne-10,000 iingxowa ze-audio ze-birdsong. Njengoko kuxhomekeke ku- inqaku, le iingcingo ze-audio ziquka kwiinkcukacha zokusebenza nge- . train_soundscapes unlabeled Building the Classifier pseudo-labeling test_soundscapes Yintoni inkxaso ifumaneka ngaphandle a ifayile. Le kathathu ibandakanya i-set ye-test audio ebandayo ngexesha lokuthumela iziphumo ze-prediction kwi-BirdCLEF+ 2025 competition. test_soundscapes readme.txt Ukwakhiwa kweClassic Umgangatho Basic kunye ne-Background Isisombululo esisiseko esetyenziswa Ukuqeqesha i-vocalization ye-bird ye-classifier yayo kubandakanya: Ukuhlolwa kwe-Google Ukuqhathanisa i-audio yokufaka kwi-5 iingxaki ze-second. Ukuguqula i-audio segments kwi-mel spectrograms. Ukuqeqesha i-image classifier kwi-mel spectrograms. Umgangatho efanayo uya kulandelayo kule nqakraza. I-image classifier eyenziwa ku-Google Model. Ukuba unemibuzo kunye iimodeli yobugcisa, uyazi ukuba ziye ziye zenzelwe ukusetyenziswa kwe-image efanelekileyo. Umgangatho B0 EfficientNet Umgangatho B0 Nangona kunjalo, phambi kokuba amaxabiso ze-audio ziyafumaneka kunye nokuguqulwa kwi-spectrograms ye-mel, kufuneka sincoma i-class imbalance kunye ne-human annotation problems ezaziwayo kwi-Mel. Ngokubanzi, iingxaki zayo ziya kuthathwa ngokufanayo nge-data augmentation kunye ne-slicing ye-audio samples. Training Data Kwiintsuku ezidlulileyo, i-sub-section ezilandelayo inikeza ulwazi olutsha olutsha. Imodeli EfficientNet I-Google Research ibonise i-family yayo iimveliso ngo-2019 njenge-set of iimveliso ezihlangene iimveliso ze- state-of-the-art, ngexesha elidlulileyo, ngokutsho ubungakanani kunye nokusebenza. EfficientNet convolutional neural network iimveliso, ezakhiwe ngo-2021, zinikeza ukusebenza nangaphezulu kunye ne-parameter efficiency. Ukucaciswa2 Nangona uqeqesho Iimodeli ze-EfficientNet zibonise izisombululo zabo xa ziye zithunyelwe kwiimveliso ezininzi zebhanki zebhanki, okwenza i-choice engundoqo njenge-technology ye-classification yeeprojekthi. Ukucinga I-Mel Spectrograms I-mel spectrogram ye-visual representation ye-audio signal. Oku kunokwenzeka ukuba iyiphi i-heatmap ye-sound. I-x-axis ye-spectrogram ye-mel ibonisa i-time dimension ye-audio signal, kwaye i-y-axis ibonisa i-frequencies yeengxaki kwi-signal. Nangona kunjalo, ngaphandle kokubonisa zonke i-frequencies kwi-scale epheleleyo, i-frequencies zihlanganiswa kwi . Lezi band, ngoko ke, zithunyelwe ukusetyenziswa . I-mel scale i-a isilinganiselo esivela kwinkqubo ye-auditory yabantu kunye neendlela yokufunda i-sound. Imibala ye-spectrogram ye-mel ibonisa i-amplitude yeengxube kwi-bands. Imibala eluhlaza ibonisa i-amplitudes ezininzi kwaye imibala eluhlaza ibonisa i-amplitudes ezincinane. mel bands mel scale logarithmic Ukucinga Indawo yami ekubuyiselwa umklamo kubonelela ukubuyekeza kwinqanaba eliphezulu kwinkqubo ngaphandle kokufumana iinkcukacha ezininzi. I-training (i-fine-tuning) logic esekelwe kule ("i-training notebook") ebandakanya kwi-4 iinkalo eziphambili: I-Notebook ye-Kaggle Isigaba 1: Ukulungiselela idatha ye-audio. Isigaba 2: Ukusetyenziswa kwedatha ye-audio. I-Section 3: Ukwenza i-spectrogram yeMel kunye nokwenza i-input. Isigaba 4: Model Ukuqeqeshwa. Ingaba uya kufumana ukuba iiyunithi ezimbini ezimbini yeenkcukacha zokuqala ze- (1) iimporti ezisetyenziselwa le nkcukacha kunye ne- (2) i-a cell defining constants ezisetyenziswa kwisiqingatha elandelayo kunye neengxenye ezidlulileyo. Config I-training notebook ngokwenene kuqala apho iipakheji base yePython ezisetyenziswa kulo lonke i-notebook zithunyelwe. Le nkcukacha kubandakanya i-logic ye-logic ye-login ("WandB") ukulawula ukuqeqeshwa. Uya kufuneka uqhagamshele WandB yakho I-notebook ye-A Ukusetyenziswa kweName . Section 0 Weights & Biases API key Kaggle Secret WANDB_API_KEY Njengoko kuxhomekeke kwi inqaku, i-unlabeled training soundscapes ingasetyenziselwa idatha ze-training nge-pseudo-labeling. Ukusetyenziswa kwe-pseudo-labeled data kubhalwe kwi Qaphela ukuba iimeko ze-Kaggle ze-non-GPU ziquka kwi-30 GiB yeememori. Training Data Section 3.5 - Pseudo-Labeling Iimodeli ebandayo emva kokufaka kwe-experimental ebandakanya kwi-sub-section ezilandelayo ifumaneka kwi-Kaggle apha. Ukuba ufuna, unako usebenzisa le model ngaphandle kokufunda yakho yayo kwaye uqhagamshelane ngqo kwi-Running Inference i-section ukuqhuba i- inference kwi-birdsong audio. Iimodeli eyenziwe emva kokufaka kwe-experimental eyenziwa kwi-sub-section ezilandelayo ifakwe kwi-Kaggle Ukuba ufuna, ungasebenzisa le model ngaphandle kokufunda yakho kunye nokufuka ngqo kwi inqaku ukuqhuba i-inference kwi-birdsong audio. Yiba Running Inference Isigaba 1 - Ukulungiswa kweedatha ye-audio Yintoni Umzekelo weNotebook: Audio Data Loading Izixhobo zihlanganisa iindidi ezininzi kwi-BirdCLEF+ 2025 umgangatho we-competition data that are not covered by the GBV classifier. Izixhobo ze-audio ezisetyenziswa nge-load_training_audio method. Yenza i-processed_audio directory kwaye ibhalisele i-copy yeedatha ye-audio ebhalisiwe njengezifayile ze-wav kwi-director. Yintoni cell of this section kuquka a I-constante. Le ngqaku ibonise inani elikhulu lwezifayili ze-audio ezivela kwi-class eyodwa. Le ngqaku iboniswe ngempumelelo kwi-value enkulu ye-audio ukuqinisekisa ukuba zonke iifayile ze-audio zithunyelwe kwi iindidi. Ingaba kufuneka uqhagamshelane le ngqongqo kwi-setting yakho ye-experimental. Ngokwesibonelo, ukuba uqhagamshelane idatha ye-audio kwizilwanyana, ungenza ukucacisa le nkqu kwi-value ephantsi ukuze akufanele ukuchithwa kwe-memory eyenziwa. Config MAX_FILES 1000 non-GBV all Yintoni indlela kungenziwa nge a i-parameter, nto leyo i-lists yeeyunithi eziquka i-audio. Kule-project, i iindidi zithunyelwe njenge-list kwaye zithunyelwe kwi-variable Kwiimeko ezidlulileyo kwi- Umgangatho nge Ukucaciswa load_training_audio classes non-GBV missing_classes load_training_audio classes # `missing_classes` list ['1139490', '1192948', '1194042', '126247', '1346504', '134933', '135045', '1462711', '1462737', '1564122', '21038', '21116', '21211', '22333', '22973', '22976', '24272', '24292', '24322', '41663', '41778', '41970', '42007', '42087', '42113', '46010', '47067', '476537', '476538', '48124', '50186', '517119', '523060', '528041', '52884', '548639', '555086', '555142', '566513', '64862', '65336', '65344', '65349', '65373', '65419', '65448', '65547', '65962', '66016', '66531', '66578', '66893', '67082', '67252', '714022', '715170', '787625', '81930', '868458', '963335', 'grasal4', 'verfly', 'y00678'] Uyakwazi ukulanda zonke iindidi ze-206 ze-BirdCLEF+ 2025 ngokuthumela iindidi eluhlaza njenge-parameter yeindidi. Uyakwazi ukulanda zonke iindidi ze-206 ze-BirdCLEF+ 2025 ngokuthumela iindidi eluhlaza njenge-parameter yeindidi. Umgangatho we-load_training_audio uyakwazi ukufumana i-use_slice ye-boolean ye-optional. Le parameter isebenza kunye ne-LOAD_SLICE constant ebonakalayo kwiseli le-Config. I-use_slice parameter kunye ne-LOAD_SLICE constant ayisetyenziselwa lo mveliso. Nangona kunjalo, ingasetyenziselwa ukulayisha inani elifanelekileyo le-audio ukusuka kwifayile ngalinye. Umzekelo, ukuba ukulayisha iiveki ezini kuphela le-audio ukusuka kwifayile ye-audio, uqhagamshelane i-LOAD_SLICE kwi-160000, ebonakalisa njengoko iiveki ezininzi le-sampling ye-32,000; kwaye uqhagamshelane yi-use_ Yintoni Method ekubeni i-optional boolean i-parameter. Le i-parameter isebenza nge- I-Constant defined kwi iimveliso Iiparamitha kunye Ukulungiselela ezisetyenziselwa le nophuhliso. Nangona kunjalo, ingasetyenziselwa ukulayisha inani elifanelekileyo ye-audio ukusuka kwifayile omnye. Umzekelo, ukuba ukulayisha kuphela iiveki ezili-5 ye-audio ukusuka kwifayile omnye we-audio, Ukucinga , leyo ilebulwe njenge times I-sampling rate ye • Ukusuka Ukusuka Ukucaciswa load_training_audio use_slice LOAD_SLICE Config use_slice LOAD_SLICE not LOAD_SLICE 160000 5 32000 True use_slice Yintoni Method akhawunti boolean i-parameter. Xa le i-parameter I-Logic yenza a isixhobo kunye nokuphepha i-copy ye-audio sample ye-audio iifayile kwi-directory. Iifayile ze-audio zithunyelwe kwi-sub-directories ezibonisa i-class eyenza. I-directory isetyenziswa kwi-section elandelayo ukugcina amaxabiso ze-audio kwi-disk ngaphandle kokuphendula amaxabiso ze-dataset ze-BirdCLEF+ 2025. load_training_audio make_copy True processed_audio .wav processed_audio Yintoni i-methode ivumela i-dictionary yeedatha ze-audio ezisetyenziswa ngokusebenzisa i-class names njenge-keys. Yonke ixabiso kwi-dictionary i-lists ye-tuples ye-form : load_training_audio (AUDIO_FILENAME, AUDIO_DATA) {'1139490': [('CSA36389.ogg', tensor([[-7.3379e-06, 1.0008e-05, -8.9483e-06, ..., 2.9978e-06, 3.4201e-06, 3.8700e-06]])), ('CSA36385.ogg', tensor([[-2.9545e-06, 2.9259e-05, 2.8138e-05, ..., -5.8680e-09, -2.3467e-09, -2.6546e-10]]))], '1192948': [('CSA36388.ogg', tensor([[ 3.7417e-06, -5.4138e-06, -3.3517e-07, ..., -2.4159e-05, -1.6547e-05, -1.8537e-05]])), ('CSA36366.ogg', tensor([[ 2.6916e-06, -1.5655e-06, -2.1533e-05, ..., -2.0132e-05, -1.9063e-05, -2.4438e-05]])), ('CSA36373.ogg', tensor([[ 3.4144e-05, -8.0636e-06, 1.4903e-06, ..., -3.8835e-05, -4.1840e-05, -4.0731e-05]])), ('CSA36358.ogg', tensor([[-1.6201e-06, 2.8240e-05, 2.9543e-05, ..., -2.9203e-04, -3.1059e-04, -2.8100e-04]]))], '1194042': [('CSA18794.ogg', tensor([[ 3.0655e-05, 4.8817e-05, 6.2794e-05, ..., -5.1450e-05, -4.8535e-05, -4.2476e-05]])), ('CSA18802.ogg', tensor([[ 6.6640e-05, 8.8530e-05, 6.4143e-05, ..., 5.3802e-07, -1.7509e-05, -4.8914e-06]])), ('CSA18783.ogg', tensor([[-8.6866e-06, -6.3421e-06, -3.1125e-05, ..., -1.7946e-04, -1.6407e-04, -1.5334e-04]]))] ...} Umgangatho uqhagamshelane i-statistics esisiseko ebonakalayo iinkcukacha zithunyelwe ngalinye iiklasi njenge-comma-separated-value string. Uyakwazi ukuhambisa iinkcukacha ngokufanelekileyo ukuze ufake iinkcukacha. class,sampling_rate,num_files,num_secs_loaded,num_files_loaded 1139490,32000,2,194,2 1192948,32000,4,420,4 1194042,32000,3,91,3 ... Section 2 - Audio Data Processing Yintoni Umzekelo weNotebook: Audio Data Processing Ukukhangisa iingxaki ze-silent kunye neengxaki ze-audio ukunciphisa iingxaki ze-humane ezininzi kwi-audio. Ukukhangisa iingxaki ze-silent ukunciphisa iingxaki zeengxaki ze-audio. I-Audio Augmentation ifumaneka (1) ukongeza isignali ye-noise esebenzayo, (2) ukuguqulwa kwe-tempo ye-ray audio, okanye (3) ukongeza isignali ye-noise esebenzayo kwaye ukuguqulwa kwe-tempo ye-ray audio. Isigaba 2.1 - Ukubonisa i-Silenced Segments Yintoni indlela isetyenziselwa "ukukhanyisa" phezu ngalinye iisampuli ye-audio ebonakalayo kunye nokukhanyisa iingxaki ze-silent ngokubala iisampuli ze-audio value of a given segment to a specified threshold. If the RMS is below the threshold, the segment is identified as a silent segment. The following constants specified in the cell of this section ukulawula isebenziswano Umgangatho: detect_silence root-mean square (RMS) Config detect_silence SIL_FRAME_PCT_OF_SR = 0.25 SIL_FRAME = int(SR * SIL_FRAME_PCT_OF_SR) SIL_HOP = int(1.0 * SIL_FRAME) SIL_THRESHOLD = 5e-5 SIL_REPLACE_VAL = -1000 # Value used to replace audio signal values within silent segments The iimveliso constants can be modified to adjust how the method "slides" over the raw audio. Similarly, the ixabiso ingasetyenziswa ukuze kwenziwe i-methode enzima okanye enzima kunceda ukucacisa iingxaki ze-silent. SIL_FRAME SIL_HOP SIL_THRESHOLD The method outputs a dictionary of silent segment markers for each file in each class. Audio files with no detected silent segments are identified by empty lists. {'1139490': {'CSA36389.ogg': [0, 8000, 16000, 272000, 280000, 288000, 296000, 304000], 'CSA36385.ogg': [0, 8000, 16000, 24000, 240000, 248000, 256000]}, '1192948': {'CSA36388.ogg': [0, 8000, 16000, 24000, 256000, 264000, 272000, 288000], 'CSA36366.ogg': [0, 8000, 16000, 24000, 256000, 264000, 272000, 280000, 288000], 'CSA36373.ogg': [0, 8000, 16000, 24000, 256000, 264000, 272000, 288000], 'CSA36358.ogg': [8000]}, '1194042': {'CSA18794.ogg': [], 'CSA18802.ogg': [], 'CSA18783.ogg': [0, 8000, 16000, 24000, 600000, 608000, 616000]}, '126247': {'XC941297.ogg': [], 'iNat1109254.ogg': [], 'iNat888527.ogg': [], 'iNat320679.ogg': [0], 'iNat888729.ogg': [], 'iNat146584.ogg': []}, '1346504': {'CSA18803.ogg': [0, 8000, 16000, 24000, 3000000, 3008000, 3016000], 'CSA18791.ogg': [], 'CSA18792.ogg': [], 'CSA18784.ogg': [0, 8000, 16000, 1232000, 1240000, 1248000], 'CSA18793.ogg': [0, 8000, 16000, 24000, 888000]} ...} Section 2.2 - Removing Silent Segments and Eliminating Human Annotations The I-Constant defined kwi cell of this section specifies if audio should be stripped of silent segments sliced to remove human annotations. USE_REMOVE_SILENCE_AND_HUMAN_ANNOT Config iimveliso Ukucinga USE_REMOVE_SILENCE_AND_HUMAN_ANNOT = True The i-methode iveza iingxaki ze-silent ezivela kwi-audio samples usebenzisa i-output method. Further, it implements logic to handle human annotations based on a simple observation: many audio samples, namely those with human annotations, to have the following structure: remove_silence_and_human_annot detect_silence tend | < 10s | ~1s | | | BIRDSONG | SILENCE | HUMAN ANNOTATION | I-birdsong kunye ne-human annotation sections ngokufanelekileyo ziquka iingxaki ze-silent. Nangona kunjalo, njengoko kubonisa kwi-diagram phezulu, iingxaki ze-bird vocalization ziyafumaneka kwimizuzu ezimbini ezidlulileyo ze-audio. Ngoko ke, indlela elula, ukuba ingxaki ze-human annotations zihlanganisa iisampuli ze-audio kwi-first silent segment marker eyenza ngaphandle kwicandelo ebonakalayo, phantsi kwimeko ukuba i-anotation ye-humane ibandakanya le-silent segment. logic uses the constant in the ifayile ukuyifaka ukuba umnqweno we-segment ye-silent ibekwe ngaphandle kwibhokisi ebonakalayo , ifumaneka kwinqanaba le-seconds. Ukuba yinto, i-logic iveza i-audio ebonakalayo kwi-markers yayo kwaye ibekwe kuphela idatha ebonakalayo phambi kokuba. Ukuhlolwa kwimifanekiso ye-audio ebonakalayo ngexesha lokucwangcisa iye yaziwa ukuba yinkqubo efanelekileyo. Nangona kunjalo, njengoko ibekwe kwi inqaku, kukho audio recordings where the human annotation the birdsong recording. The logic described here does Zonke iimveliso ze-audio zihlanganisa iinkqubo ezininzi ze-birdsong ezihambisiweyo kunye neemveliso ezininzi ziquka iingxaki ze-silent. Iimveliso zezi zihlanganiswa ngexesha elandelayo kwaye zihlanganiswa ngokupheleleyo. remove_silence_and_human_annot ANNOT_BREAKPOINT Config ANNOT_BREAKPOINT Training Data Zonke precedes not Iingubo yesibini esisodwa, , can be optionally used in a final processing step to return an even more refined slice of the processed audio. Set Imininingwane yeenkcukacha ze-audio ezisetyenziswa ukuba ushiye. SLICE_FRAME SLICE_FRAME Yintoni Method ukugcina i-audio eyenziwe kwi-disk phantsi kwe-directory Ukusebenzisa i I-parameter, ebizwa ngokuba yi- . Umgangatho ivumela i-Dictionary of the seconds of processed audio for each class. remove_silence_and_human_annot processed_audio save_audio True Ukubalwa {'1139490': 14, '1192948': 29, '1194042': 24, '126247': 48, '1346504': 40, '134933': 32, '135045': 77, ...} Yintoni Uhlobo olusetyenziswa emva ukuze ufumane inani elidlulileyo lombala ye-audio kwiiyunithi ezininzi. get_audio_stats remove_silence_and_human_annot Section 2.3 - Calculating Augmentation Turns for Minority Classes Njengoko kulandwa kwi section, the classes are not balanced. Augmentation is used in this notebook section to help address the imbalance leveraging the average number of seconds of audio across all classes, as provided by the Method. Classes nge-seconds epheleleyo ye-audio eyakhelwe the average are augmented. The i-methode ibonise inani le-augmentation turns ngalinye i-minority class ngokusebenzisa inani le-seconds ngalinye i-processed audio sample. Training Data get_audio_stats below get_augmentation_turns_per_class TURNS = (AVG_SECS_AUDIO_ACROSS_CLASSES - TOTAL_SECS_AUDIO_FOR_CLASS)/AVG_SECS_PER_AUDIO_SAMPLE Iiyunivesithi ezininzi phantsi komgangatho iya kuba i-augmentation turns ngaphezulu kunokuba iiyunivesithi ezininzi ziquka komgangatho, leyo iya kuba i-augmentation turns ezincinane. The includes a constant which can be used to adjust the value for get_augmentation_turns_per_class AVG_SECS_FACTOR average number of seconds of audio across all classes. The constant can be used to make the logic more conservative or aggressive when calculating the number of augmentation turns. Yintoni Ukubandakanya a umgangatho, leyo ingasetyenziswa ukuguqulwa kwinqanaba get_augmentation_turns_per_class AVG_SECS_FACTOR ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane ubuncinane. Section 2.4 - Running Augmentations The constant defined in the i-cell ye-section inikeza ukuba i-audio kufuneka ifumaneka. USE_AUGMENTATIONS Config USE_AUGMENTATIONS = True Njengoko kwangaphambili, i-audio augmentation ibandakanya (1) ukongeza isignali ye-noise eyenziwe ngempumelelo, (2) ukuguqulwa kwe-tempo ye-ray audio, okanye (3) ukongeza isignali ye-noise eyenziwe ngempumelelo kwaye ukuguqulwa kwe-tempo ye-ray audio. iimveliso methods encapsulate the logic for adding a noise signal and changing the tempo respectively. The noise signal range and tempo change range can be adjusted via the following constants in the Ukucinga: add_noise change_tempo Config NOISE_RNG_LOW = 0.0001 NOISE_RNG_HIGH = 0.0009 TEMPO_RNG_LOW = 0.5 TEMPO_RNG_HIGH = 1.5 The method runs the augmentations using the output from the indlela. Kuba izifundo ezidlulileyo, i-logic: run_augmentations get_augmentations_turns_per_class Ukukhetha ngempumelelo isampula ye-audio eyenziwe (i.e. iingxaki ze-silent ziye zithunyelwe) ukuze ifumaneke. Randomly selects the augmentation to perform: (1) adding noise, (2) changing the tempo, or (3) adding noise and changing the tempo. Ukugcina i-audio ephakamileyo kwi-disk phantsi kwe-class efanelekileyo kwi-processed_audio directory. Nangona i-notebook logic ibandakanya iiyunithi ezininzi kunye neentsuku ezininzi ze-audio phantsi komgangatho, ibandakanya iiyunithi ezininzi kunye neentsuku ezininzi ze-audio phantsi komgangatho. Le nqakraza yenzelwe ukulawula i-memory ezikhoyo kunye nokufumaneka ukuba i-class imbalance ibandakanya ngakumbi ngokusebenzisa ukhetho lwe-loss function. Section 3 - Mel Spectrogram Generation and Input Preparation The section of the notebook: Mel Spectrogram Generation and Input Preparation Ukuqhathanisa idatha ye-audio eyenziwe kwi-training kunye ne-validation lists. Splits audio into 5 second frames. Yenza i-mel spectrograms ngalinye i-frame ye-audio ye-5 ye-second. Resizes mel spectrograms to a target size of . (224, 224) Optionally loads pseudo-labeled data samples to augment training data. One-hot encodes training data and validation data labels. Ukwakha i-TensorFlow Dataset i-objects kwi-training kunye ne-validation data lists. Optionally uses MixUp logic to augment training data. Section 3.1 - Splitting Processed Audio Data Processed audio data is loaded from the folder. The data is split into 4 lists: processed_audio training_audio training_labels validation_audio validation_labels I-Etiquette, ngoko ke, iimveliso ze-class ezinxulumene neempawu ze-audio. I-Etiquette constant defined in the cell controls the split ratio between the training and validation data lists. Processed audio data is shuffled before splitting. SPLIT Config I-Section 3.2 - Ukubala i-Audio kwi-frames I-Audio ifakwe kwi-5 segments ye-second usebenzisa i indlela, nto leyo isebenzisa i-TensorFlow umgangatho ukwahlula umzekelo we-audio. I-constants ezilandelayo kwi cell control the split operation: frame_audio signal.frame Config FRAME_LENGTH = 5 FRAME_STEP = 5 Section 3.3 - Generating Mel Spectrograms Mel spectrograms are generated for each 5 second audio frame generated in Ukusebenzisa i indlela. Iimeko ezilandelayo kwi cell kucacisa iiparamitha ezisetyenziswa ekwakheni i-mel spectrograms, njenge-number of mel bands, i-frequency engaphansi, kunye ne-frequency engaphezulu: Section 3.2 audio2melspec Config # Mel spectrogram parameters N_FFT = 1024 # FFT size HOP_SIZE = 256 N_MELS = 256 FMIN = 50 # minimum frequency FMAX = 14000 # maximum frequency I-band ye-frequency ilungiselwe ukuba ibonisa i- iindidi ezininzi ze-bird vocalizations. Nangona kunjalo, ezinye iintlobo ze-bird zinokufunda ngaphandle kwesixeko. Umthamo Isigaba 3.4 - Ukuguqulwa kwe-Mel Spectrograms Yintoni indlela isetyenziselwa ukuguqulwa yonke i-mel spectrogram kwi-a Objects. Zonke object is subsequently resized to Yintoni i-input dimension esithathwe yi-EfficientNet B0 model. to_melspectrogram_image pillow Image Image (224, 224) Section 3.5 - Loading Pseudo-Labeled Data Njengoko kulandwa kwi section, the inkxaso ibandakanya malunga ne-10,000 audio recordings of birdsong. These audio recordings can be incorporated into the training data via . A simple process to create pseudo-labeled data is as follows: Training Data train_soundscapes unlabeled pseudo-labeling Train a classifier pseudo-labeled data. without Load training soundscape audio files. I-Segmentation ye-audio ye-soundscape kwi-frames ye-5 ye-second. Yenza i-mel spectrograms ngalinye i-frame ye-5 ye-second kunye nokuguqulwa kwe-224, 224. Yenza imibuzo kwi-spectrogram ye-mel ye-re-size usebenzisa i-classifier eyenziwe ngexesha lokuqala. Hlola iingcebiso ngaphezu kwinqanaba elidlulileyo yobugcisa kunye nokugcina i-spectrograms yeengcebiso zihlanganisa phantsi kwe-label ye-class ebonakalayo. Train your classifier again using the psuedo-labeled data. Ukuba ufuna ukuvelisa idatha yakho ye-pseudo-labelled, uqhagamshelane neengxaki ezininzi ukuvelisa i-classifier yakho iinkcukacha ze-pseudo-labelled. Emva koko, usebenzisa i-classizer yakho ukuvelisa iinkcukacha zakho ze-pseudo-labelled usebenzisa inkqubo efumaneka phezulu. Ekugqibeleni, uphuhlise i-classizer yakho usebenzisa iinkcukacha zakho ze-pseudo-labelled. ngaphandle Ukusetyenziswa oku kukusebenzisa iinkcukacha ze-pseudo-labeled. Nangona kunjalo, unako ukuguqulwa i-inference notebook ebizwa inqaku ukuvelisa idatha ye-pseudo-labeled. Running Inference Yenza i Ukulungiselela kwi I-Cell 2 ukunceda ukusetyenziswa kwedatha ze-pseudo-labeled. USE_PSEUDO_LABELS Config False Section 3.6 - Encoding Labels Yintoni indlela isetyenziselwa i-one-hot encode labels. I-one-hot encoded labels zithunyelwe njenge-NumPy array kwaye ziquka kwi-training label kunye ne-validation label lists. process_labels I-Section 3.7 - Ukuguqulwa kwe-Training and Validation Data Lists kwi-TensorFlow Iimpawu Dataset Dataset I-TensorFlow indlela isetyenziselwa ukuvelisa TensorFlow izixhobo ze-training kunye ne-validation data lists. The indlela ibizwa kwi-training Objection ukuba shuffle idatha yokulungisa ngaphambi batching. The Uhlobo olusetyenziswa kumabini objects to batch the training and validation datasets. The Ukulungiselela kwi cell ukulawula ubungakanani batch. data.Dataset.from_tensor_slices Dataset shuffle Dataset batch Dataset BATCH_SIZE Config I-Section 3.8 - Ukusebenzisa i-MixUp ukwandisa i-Training Data Njengoko uyazi xa, i-MixUp yinkqubo yokwandisa idatha enokufaka iifoto ezimbini kunye nokwenza isampuli entsha yeedatha. I-class ye-imaging ye-blended is a mixture of the classes associated with the original 2 images. indlela, kunye ne method, encapsulates the optional MixUp logic. mix_up sample_beta_distribution Ukusetyenziswa oku kusebenzisa i-MixUp ukwandisa idatha ye-training. Ukusetyenzisa i-MixUp, Ukulungiselela kwi I-Cell 2 . USE_MIXUP Config True Isigaba 4 - Model Training Yintoni Umzekelo weNotebook: Model Training I-Inicializes kunye ne-configures yeprojekthi ye-WandB ukuthatha idatha ze-training run. Ukwakhiwa kunye nokwakhiwa kwimodeli EfficientNet B0. Ukuqeqesha imodeli. Yenza iimodeli ezidlulileyo kwi-disk. Section 4.1 - Initializing and Configuring WandB Project Qinisekisa ukuba uqhagamshelane key yakho WandB API njenge Kaggle Secret kwi-notebook kwaye ukuba i-WandB login method kwi-Section 0 ye-notebook iye yenza True. Ensure that you have attached your own WandB API key as a kwi-notebook kunye ne-WandB Ukucinga I-notebook yaye yaye yibhalwe . Kaggle Secret login Section 0 True Yintoni iseli kule ndawo ibandakanya i-logic yokufaka kunye nokufaka inkqubo entsha ye-WandB (ukuba inkqubo ayikho ngoku) leyo uya kufumana idatha ze-training run: Config wandb.init(project="my-bird-vocalization-classifier") config = wandb.config config.batch_size = BATCH_SIZE config.epochs = 30 config.image_size = IMG_SIZE config.num_classes = len(LABELS) Kwakhona, unako ukuguqulwa igama le nkqubo ukuba igama lakho leprojekthi WandB efunyenwe. my-bird-vocalization-classifier Isigaba 4.2 - Ukwakhiwa kunye nokwakhiwa Model EfficientNet B0 Yintoni indlela isetyenziselwa ukulayisha imodeli efanelekileyo EfficientNet B0 kunye ne-ImageNet we-weights kwaye ngaphandle kwe-top layer: build_model model = EfficientNetB0(include_top=False, input_tensor=inputs, weights="imagenet") The model is frozen to leverage the pre-trained ImageNet weights with the objective to only (i.e. isitimela) izihlangu kwi-stage yokuqala ye-model: Ukucinga # Unfreeze last `unfreeze_layers` layers and add regularization for layer in model.layers[-unfreeze_layers:]: if not isinstance(layer, layers.BatchNormalization): layer.trainable = True layer.kernel_regularizer = tf.keras.regularizers.l2(L2_RATE) The constant Kwimeko cell ibonise inani layers ukuba unfreeze. UNFREEZE_LAYERS Config The top of the model is rebuilt with a final layer reflecting the number bird species classes. ifumaneka njenge-Loss Function ukunceda ukuguqulwa kwe-class imbalance. I iimveliso I-Constants kwi cells zisetyenziswa kunye ne-function loss. Dense Categorical focal cross-entropy LOSS_ALPHA LOSS_GAMMA Config Isigaba 4.3 - Model Training Yintoni method is called on the compiled ukusuka ukuqhuba ukuqeqeshwa. Qaphela ukuba a iimveliso , isetyenziswa ngempumelelo yokufunda. Isilinganiso yokufunda yokufunda i-callback ye-hardcoded. I-learning rate ifumaneka kwiiyure ezimbini ezisekusene ne-epoch count. Inani le-epoch ye-training ilawulwa ngu-epoch Ukulungiselela kwi Ukucinga fit model Section 4.2 learning rate scheduler lr_scheduler 4.0e-4 EPOCHS Config Isigaba 4.4 - Model Saving Yintoni indlela ibizwa ku-compiled emva kokufunda ukugcina imodeli kwi-disk. save model model.save("bird-vocalization-classifier.keras") Imiphumo yemfundo Ukusebenza kwe-notebook kufuneka yenza imiphumo ye-training ezilandelayo, ngexesha ukuba usebenzise isakhiwo se-experimental esifundowe kwisisombululo. Ukucaciswa: Building the Classifier Njengoko kubonakala, ukucaciswa kwama-90% kwaye ukucaciswa kwama-70% emva kokufunda kwama-30 epochs. Nangona kunjalo, njengoko kubonakala, ukucaciswa kwama-validation kubaluleke kakhulu. Le ingxaki kubaluleke ngexabiso se-class kunye ne-memory eyenza ukusetyenziswa kwezongezelelo ezongezelelweyo ukucaciswa ngokupheleleyo. Iimiphumo zibonisa ukuba iimodeli ivimbele ukusetyenziswa kwedatha ze-training kwaye ayikho ukuguqulwa njengoko kunokufuneka. Nangona kunjalo, iimodeli ingasetyenziselwa ukucaciswa kunye ne-GBV ngokufanelekileyo ngokufanelekileyo. Ukusuka Inference Yintoni ("inference notebook") can be used for running inference. The inference notebook logic uses both the GBV classifier model and the model that you trained in the preceding section. It runs inference on the unlabeled soundscapes files in the Iifayile ye-soundcapes ye-audio yahlukaniswa kwi-5 iifayile ze-second. I-soundcapes ye-audio I-Constant defined kwi cell of of the notebook ukulawula inani iifayile ye-soundcapes ye-audio eyenziwa ukususela. I-Notebook ye-Kaggle train_soundscapes MAX_FILES Config Section 0 I-Inference Notebook kuqala ukuvelisa imibuzo usebenzisa i-GBV classifier. Imibuzo ye 143 iindidi ze-BirdCLEF+ 2025 ezihlangene kwi-GBV classifier yaziwa. Ukuba i-probability ephakeme phakathi kwe-143 iindidi ezaziwayo iye aphezulu okanye efana ne: , ngoko i-GBV ebonakalayo i-class ifumaneka njenge-class efanelekileyo. Ukuba i-probability ephakeme phakathi kwe-143 "ezaziwayo" i-class ifumaneka phantsi , it is assumed that the true class is among the 63 classes "unknown" to the GBV classifier - i.e. the classes used to train the model in the preceding section. The logic then runs predictions using the finetuned model. The predicted class from that prediction set is subsequently selected as the true class. GBV_CLASSIFIER_THRESHOLD GBV_CLASSIFIER_THRESHOLD Yintoni I-Constante ifumaneka kwi- I-Cell ye of the inference notebook. Izixhobo zithunyelwe kwiifayile ze-2: GBV_CLASSIFIER_THRESHOLD Config Section 5 A file that captures the prediction and prediction probability for each 5-second soundscape slice. preds.csv Iifayile ye-submission.csv enikeza zonke iimeko ze-class kwi-format ye-BirdCLEF+ 2025. Yenza i-path to your finetuned model kwi-cell yokuqala ye-Section 4 ye-Inference Notebook. Yenza i-path to your finetuned model kwi-cell yokuqala Umbhali we-Inference Notebook Section 4 Future Work I-training notebook ingasetyenziselwa ukuqeqesha umzila kwi-206 BirdCLEF+ 2025 izifundo, ukunciphisa inkinga ye-GBV classifier, ngexabiso malunga ne-concours data set. Njengoko kwangaphambili, ukufumana isifundo esifutshane, Ukusuka ku umgangatho uqhagamshelane idatha ye-audio evela kumazwe omnye. I iimveliso constants can be used to limit the amount of loaded audio in order to work within the confines of a Kaggle notebook environment. [] load_training_audio MAX_FILES LOAD_SLICE Nangona kunjalo, iimodeli efanelekileyo ingasetyenziselwa ukusetyenziswa kwamathambo emininzi yobugcisa. Kwiimeko ezininzi, inani elikhulu le-augmentations iya kuthathwa ukunyuswa kwizinga le-class. Ukongezelela, ezinye izixhobo ze-augmentation, njenge-CutMix, ingasetyenziselwa ukunyuswa kwakhona iidatha ze-training. Nangona kunjalo, lezi zindlela zinxibelelana kwimeko yokukhula emangalisayo.