RESUMO
We investigate the impact of different data modalities for cattle weight estimation. For this purpose, we collect and present our own cattle dataset representing the data modalities: RGB, depth, combined RGB and depth, segmentation, and combined segmentation and depth information. We explore a recent vision-transformer-based zero-shot model proposed by Meta AI Research for producing the segmentation data modality and for extracting the cattle-only region from the images. For experimental analysis, we consider three baseline deep learning models. The objective is to assess how the integration of diverse data sources influences the accuracy and robustness of the deep learning models considering four different performance metrics: mean absolute error (MAE), root mean squared error (RMSE), mean absolute percentage error (MAPE), and R-squared (R2). We explore the synergies and challenges associated with each modality and their combined use in enhancing the precision of cattle weight prediction. Through comprehensive experimentation and evaluation, we aim to provide insights into the effectiveness of different data modalities in improving the performance of established deep learning models, facilitating informed decision-making for precision livestock management systems.
RESUMO
To make timing of artificial insemination (AI) relative to ovulation less critical, methods for prolonging shelf life of spermatozoa in vivo after AI have been attempted to be developed. Encapsulation of sperm cells is a documented technology, and recently, a technology in which sperm cells are embedded in alginate gel has been introduced and commercialized. In this study, standard processed semen with the Biladyl extender (control) was compared with semen processed by sperm immobilization technology developed by SpermVital AS in a blind field trial. Moreover, in vitro acrosome and plasma membrane integrity was assessed and compared with AI fertility data for possible correlation. Semen from 16 Norwegian Red young bulls with unknown fertility was collected and processed after splitting the semen in two aliquots. These aliquots were processed with the standard Biladyl extender or the SpermVital extender to a final number of 12 × 10(6) and 25 × 10(6) spermatozoa/dose, respectively. In total, 2000 semen doses were produced from each bull, divided equally by treatment. Artificial insemination doses were set up to design a blinded AI regime; 5 + 5 straws from each extender within ejaculates in ten-straw goblets were distributed to AI technicians and veterinarians all over Norway. Outcomes of the inseminations were measured as 56-day nonreturn rate (NRR). Postthaw sperm quality was assessed by flow cytometry using propidium iodide and Alexa 488-conjugated peanut agglutinin to assess the proportion of plasma membrane and acrosome-intact sperm cells, respectively. In total, data from 14,125 first inseminations performed over a 12-month period, 7081 with Biladyl and 7044 with SpermVital semen, were used in the statistical analyses. There was no significant difference in 56-day NRR for the two semen categories, overall NRR being 72.5% and 72.7% for Biladyl and SpermVital, respectively. The flow cytometric results revealed a significant higher level of acrosome-intact live spermatozoa in Biladyl-processed semen compared to SpermVital semen. The results indicate that the level of acrosome-intact live spermatozoa in the AI dose did not affect the 56-day NRR for the two semen processing methods. In conclusion, this study has showed that immobilized spermatozoa provide equal fertility results as standard processed semen when AI is performed in a blinded field trial, although the immobilization procedure caused increased sperm damage evaluated in vitro compared to standard semen processing procedure.
