[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84683-en":3,"doc-seo-84683-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84683,4398048949847,"Eliana","https://ap-avatar.wpscdn.com/avatar/400002536579ef2da7f?_k=1778318612642679267",8,"Research & Report","Adaptive Loss Balancing for Multi-Task Bioacoustic Classification of Bird Species and Call Types","Reliable analysis of bird vocalisations in passive acoustic monitoring requires multi-target models that handle imbalanced annotations. The work extends BirdCallNet for joint species and call-type classification on the long-tailed WiWa forest-soundscape dataset, examining how task-loss balancing interacts with pretrained representations and adaptation depth. Four bird-domain encoders are evaluated under linear probing, attentive probing, and full fine-tuning. Fixed weighting, homoscedastic uncertainty, Dynamic Weight Averaging, and GradNorm are compared.","arXiv :2607 .03304v 1 [ cs . SD] 3 Jul 2026  \nAdaptive Loss Balancing for Multi-Task Bioacoustic Classification of Bird Species and Call  \nTypes  \nParia Vali Zadeh 1[0009−0007−8396−1585] and Sven Tomforde 1[0000−0002−5825−8915]  \nKiel University, Kiel, Germany  \nparia.vali .zadeh@cs.uni-kiel .de  \n[https://www.uni-kiel.de](https://www.uni-kiel.de)  \nAbstract. Reliable analysis of bird vocalisations in passive acoustic monitoring requires models that can handle multiple, imbalanced annotation targets. We extend BirdCallNet for joint species and call-type classification on the long-tailed WiWa forest-soundscape dataset and investigate how task-loss balancing interacts with pretrained representations and adaptation depth. Four bird-domain encoders–ConvNeXtBS , EAT, BirdMAE, and ProtoCLR–are evaluated with separate species and call-type heads under linear probing, attentive probing, and full fine-tuning. A manually tuned fixed objective is compared with homoscedastic uncertainty weighting and Dynamic Weight Averaging across all three adaptation regimes, while GradNorm is evaluated only under full fine-tuning.  \nThe results indicate that the factorised multi-task formulation yields the most consistent improvements over the combined single-task baseline for call-type recognition, while its effect on species recognition depends more strongly on the adaptation regime. Full fine-tuning is not consistently optimal: ConvNeXtBS achieves the highest mean species performance under linear probing, whereas BirdMAE provides the strongest call-type performance under attentive probing. Adaptive weighting benefits species recognition more consistently than call-type recognition. Uncertainty weighting is particularly effective for species recognition under attentive probing, whereas Dynamic Weight Averaging is generally stronger for the same task under full fine-tuning. GradNorm achieves competitive call-type performance for selected backbones but consistently underperforms the other weighting strategies for species recognition, while incurring higher computational and memory costs. Overall, the preferred loss-balancing strategy depends on the backbone, adaptation regime, and target task, while frozen-backbone adaptation can provide a more favourable performance–efficiency trade-off than end-to-end fine-tuning.  \nKeywords: Bioacoustics · Passive Acoustic Monitoring · Multi-Task Learning · Adaptive Loss Balancing · Bird Species Classification · CallType Classification.  \n2 P. Vali Zadeh and S. Tomforde  \n1 Introduction  \nBird populations are increasingly affected by anthropogenic pressures, including habitat loss and fragmentation, land-use change, climate change, and mortality associated with human-made structures [19] . These pressures can alter species distributions, reduce population sizes, and increase the risk of local or global extinction [19,5] . Large-scale studies have reported substantial declines in bird abundance, further emphasising the need for reliable monitoring approaches that can track population changes and support conservation planning [40, 11 ,28] .  \nEstimating bird population status and understanding the drivers of population change remain challenging. Conventional field surveys can provide high-quality observations, but they are often labour-intensive, spatially restricted, and difficult to repeat continuously across large temporal and geographic scales [15,45] . Moreover, estimating anthropogenic mortality or population-level effects is complicated by incomplete observations, detection bias, and variation in survey conditions [28, 11] . These limitations motivate the development of complementary monitoring methods, including passive acoustic monitoring, that can extend ecological surveys across broader spatial and temporal scales [15,41] .  \nBioacoustic monitoring offers a scalable and non-invasive way to observe wildlife and acoustic ecosystems without requiring continuous human presence in the field [15,45] . Passive aco","cbCaieCPxJgDZGn2","https://ap.wps.com/l/cbCaieCPxJgDZGn2","pdf",859979,1,30,"English","en",105,"# Introduction\n## Motivation for passive acoustic monitoring\n## Challenges in automated bird sound analysis\n## Why call-type information matters\n# Adaptive loss balancing for bioacoustic classification\n## Extending BirdCallNet for multi-task learning\n## Task-loss balancing methods and evaluation regimes","[{\"question\":\"What problem does the paper address in bird vocalisation analysis?\",\"answer\":\"It targets reliable passive acoustic monitoring when models must handle multiple, imbalanced annotation targets such as bird species and call types.\"},{\"question\":\"How is BirdCallNet extended in this work?\",\"answer\":\"The paper extends BirdCallNet to jointly classify species and call types on the long-tailed WiWa forest-soundscape dataset and studies how loss balancing interacts with pretrained representations and adaptation depth.\"},{\"question\":\"Which loss-balancing approaches are compared, and under what settings?\",\"answer\":\"It compares a manually tuned fixed objective, homoscedastic uncertainty weighting, and Dynamic Weight Averaging across linear probing, attentive probing, and full fine-tuning, while GradNorm is evaluated only under full fine-tuning.\"}]",1784197646,76,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"adaptive-loss-balancing-for-multi-task-bioacoustic-classification-of-bird-species-and-call-types","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/adaptive-loss-balancing-for-multi-task-bioacoustic-classification-of-bird-species-and-call-types/84683/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does the paper address in bird vocalisation analysis?","Question",{"text":75,"@type":76},"It targets reliable passive acoustic monitoring when models must handle multiple, imbalanced annotation targets such as bird species and call types.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How is BirdCallNet extended in this work?",{"text":80,"@type":76},"The paper extends BirdCallNet to jointly classify species and call types on the long-tailed WiWa forest-soundscape dataset and studies how loss balancing interacts with pretrained representations and adaptation depth.",{"name":82,"@type":73,"acceptedAnswer":83},"Which loss-balancing approaches are compared, and under what settings?",{"text":84,"@type":76},"It compares a manually tuned fixed objective, homoscedastic uncertainty weighting, and Dynamic Weight Averaging across linear probing, attentive probing, and full fine-tuning, while GradNorm is evaluated only under full fine-tuning.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":21,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]