Zobrazit v EDS

Monitoring the ramp use of cage-free laying hens with deep learning technologies.

Uloženo v:
Podrobná bibliografie
Název: Monitoring the ramp use of cage-free laying hens with deep learning technologies.
Autoři: Paneru B; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Yang X; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Dhungana A; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Dahal S; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Ritz CW; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Kim W; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA., Liu T; School of Computing, University of Georgia, Athens, GA 30602, USA., Chai L; Department of Poultry Science, College of Agricultural & Environmental Sciences, University of Georgia, Athens, GA 30602, USA. Electronic address: lchai@uga.edu.
Zdroj: Poultry science [Poult Sci] 2025 Nov; Vol. 104 (11), pp. 105858. Date of Electronic Publication: 2025 Sep 20.
Způsob vydávání: Journal Article
Jazyk: English
Informace o časopise: Publisher: Elsevier Country of Publication: England NLM ID: 0401150 Publication Model: Print-Electronic Cited Medium: Internet ISSN: 1525-3171 (Electronic) Linking ISSN: 00325791 NLM ISO Abbreviation: Poult Sci Subsets: MEDLINE
Imprint Name(s): Publication: 2020- : [Cambridge, UK] : Elsevier
Original Publication: Champaign Il : Poultry Science Association
Výrazy ze slovníku MeSH: Chickens*/physiology , Deep Learning* , Animal Husbandry*/methods , Animal Husbandry*/instrumentation , Housing, Animal* , Nesting Behavior*, Animals ; Female ; Animal Welfare
Abstrakt: Competing Interests: Disclosures The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Mislaid eggs are management and economic challenges in Cage-free (CF) housing systems, producing about 10-15 % floor eggs, although approximately 40 % of table egg laying hens are now in CF production systems. Ramps may ease the use of nesting boxes by minimizing physical extension to hens, thereby reducing floor eggs, and increasing the number of eggs laid in the nest box. The objective of our study was to develop a deep learning method for monitoring hens' ramp use and the potential impact of ramp access to the nest box on the number of floor eggs and the number of eggs laid in the nest boxes. A total of 600 Lohmann LSL Lite hens were raised from day 1 to 413 in three identical research rooms (7.3 m L × 6.1 m W × 3 m H) following the Lohmann LSL Lite management guide. Each study room had four nest boxes placed at the four corners of the room. Two nest boxes were provided with ramp access (R), and two nest boxes had no ramp access (NR) in each study room and were replicated among three study rooms. The ramp use was video recorded at 15 frames per second (fps). We trained two You Only Look Once (YOLO) models, YOLOv5u and YOLO11, object detection models for 200 epochs each. A total of 2,000 images were used for training (70 %), validation (20 %), and testing (10 %) for the model. All models achieved a precision, recall, and mean average precision at 0.50 intersection over union (mAP@0.50) of at least 0.94. YOLO11n(nano) achieved the highest precision (0.9940), recall (0.9934), and mAP@0.50 (0.9848). Our best model provides a baseline for automatic ramp use detection with 0.99 precision. Ramp access did not lower the floor egg production statistically (p = 0.5468). Across bird weeks, ramp access (R) and no ramp access (NR) revealed opposite patterns in floor and nest-box egg production. Ramp access generally resulted in significantly higher floor egg production in several weeks (52, 54, 58), but also produced higher nest-box egg counts in other weeks (49, 51, 52, 54, 56, 58), whereas no-ramp access showed more nest-box access in weeks when ramp access floor egg were high (48, 50, 53, 55, 57, 59). Overall, ramp access did not consistently increase or decrease egg laying in either location, but shifted the proportion of eggs between the floor and nest-boxes depending on bird age (week). Future studies are warranted to investigate the effect of ramp use on nesting behavior and floor eggs from egg laying to the end of the laying cycle. Data on floor eggs and nest box eggs with ramp access to the nest box from commercial aviary systems, as well as the CF system, also need to be compared.
(Copyright © 2025. Published by Elsevier Inc.)
Contributed Indexing: Keywords: Cage-free housing; floor eggs; machine vision; nesting behavior; ramp use
Entry Date(s): Date Created: 20250925 Date Completed: 20251129 Latest Revision: 20251129
Update Code: 20251129
PubMed Central ID: PMC12508830
DOI: 10.1016/j.psj.2025.105858
PMID: 40997600
Databáze: MEDLINE
Popis
Abstrakt:Competing Interests: Disclosures The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.<br />Mislaid eggs are management and economic challenges in Cage-free (CF) housing systems, producing about 10-15 % floor eggs, although approximately 40 % of table egg laying hens are now in CF production systems. Ramps may ease the use of nesting boxes by minimizing physical extension to hens, thereby reducing floor eggs, and increasing the number of eggs laid in the nest box. The objective of our study was to develop a deep learning method for monitoring hens' ramp use and the potential impact of ramp access to the nest box on the number of floor eggs and the number of eggs laid in the nest boxes. A total of 600 Lohmann LSL Lite hens were raised from day 1 to 413 in three identical research rooms (7.3 m L × 6.1 m W × 3 m H) following the Lohmann LSL Lite management guide. Each study room had four nest boxes placed at the four corners of the room. Two nest boxes were provided with ramp access (R), and two nest boxes had no ramp access (NR) in each study room and were replicated among three study rooms. The ramp use was video recorded at 15 frames per second (fps). We trained two You Only Look Once (YOLO) models, YOLOv5u and YOLO11, object detection models for 200 epochs each. A total of 2,000 images were used for training (70 %), validation (20 %), and testing (10 %) for the model. All models achieved a precision, recall, and mean average precision at 0.50 intersection over union (mAP@0.50) of at least 0.94. YOLO11n(nano) achieved the highest precision (0.9940), recall (0.9934), and mAP@0.50 (0.9848). Our best model provides a baseline for automatic ramp use detection with 0.99 precision. Ramp access did not lower the floor egg production statistically (p = 0.5468). Across bird weeks, ramp access (R) and no ramp access (NR) revealed opposite patterns in floor and nest-box egg production. Ramp access generally resulted in significantly higher floor egg production in several weeks (52, 54, 58), but also produced higher nest-box egg counts in other weeks (49, 51, 52, 54, 56, 58), whereas no-ramp access showed more nest-box access in weeks when ramp access floor egg were high (48, 50, 53, 55, 57, 59). Overall, ramp access did not consistently increase or decrease egg laying in either location, but shifted the proportion of eggs between the floor and nest-boxes depending on bird age (week). Future studies are warranted to investigate the effect of ramp use on nesting behavior and floor eggs from egg laying to the end of the laying cycle. Data on floor eggs and nest box eggs with ramp access to the nest box from commercial aviary systems, as well as the CF system, also need to be compared.<br /> (Copyright © 2025. Published by Elsevier Inc.)
ISSN:1525-3171
DOI:10.1016/j.psj.2025.105858