Fig 1.
Parturition data collection.
Fig 2.
Display of image enhancement effect.
Table 1.
Equine parturition dataset.
Table 2.
Test set labelling details.
Fig 3.
The upper part of the CBAM architecture consists of two components: the channel attention module and the spatial attention module. These modules are applied to the input feature map, and the output is obtained by element-wise multiplication with the input feature map. The lower part shows the specific placement of the CBAM module in ResNet101, where it is added at the end of each residual unit in the third and fourth stages.
Fig 4.
The input feature maps come from the outputs of the FPN, namely P2’, P3’, P4’, and P5’. These feature maps are processed through the Integrate and Refine operations to generate new feature maps P2, P3, P4, and P5, which are used for subsequent object detection tasks.
Fig 5.
GRoIE Module.
Fig 6.
L-MPD network architecture.
Table 3.
Experimental parameter setting.
Table 4.
Evaluation results of Libra R-CNN with different backbone networks.
Table 5.
Ablation experiment.
Fig 7.
Performance evaluation and training process analysis for key components in ablation study.
Table 6.
Comparison of recognition of standing and lateral recumbent parturition in mare.
Table 7.
Different object detection algorithm performance comparison.
Fig 8.
Mare parturition identification under video stream continuous monitoring scenarios.
Table 8.
Ten-fold cross-validation.
Fig 9.
Category activation mapping.
Fig 10.
Examples of recognition errors.
(a) misidentification due to insufficient lighting, (b) indistinct features due to camera angle and capture range, (c) intrusive interference, (d) tail occlusion, (e) non-parturition mares misidentified as parturition mares due to environmental interference.