DPSA: dense pixelwise spatial attention network for hatching egg fertility detection

Abstract

© 2020 SPIE and IS & T. Deep convolutional neural networks show a good prospect in the fertility detection and classification of specific pathogen-free hatching egg embryos in the production of avian influenza vaccine, and our previous work has mainly investigated three factors of networks to push performance: depth, width, and cardinality. However, an important problem that feeble embryos with weak blood vessels interfering with the classification of resilient fertile ones remains. Inspired by fine-grained classification, we introduce the attention mechanism into our model by proposing a dense pixelwise spatial attention module combined with the existing channel attention through depthwise separable convolutions to further enhance the network class-discriminative ability. In our fused attention module, depthwise convolutions are used for channel-specific features learning, and dilated convolutions with different sampling rates are adopted to capture spatial multiscale context and preserve rich detail, which can maintain high resolution and increase receptive fields simultaneously. The attention mask with strong semantic information generated by aggregating outputs of the spatial pyramid dilated convolution is broadcasted to low-level features via elementwise multiplications, serving as a feature selector to emphasize informative features and suppress less useful ones. A series of experiments conducted on our hatching egg dataset show that our attention network achieves a lower misjudgment rate on weak embryos and a more stable accuracy, which is up to 98.3% and 99.1% on 5-day and 9-day old eggs, respectively.

An Adaptive Partial Least-Squares Regression Approach for Classifying Chicken Egg Fertility by Hyperspectral Imaging

“…These results showed that the PLS algorithm used can discriminate both brown and white fertile eggs from non-fertile eggs prior to incubation, using any of the thresholds identified. This is a huge contribution when compared to earlier studies that have not achieved such accuracies prior to incubation, with some studies reporting only the overall accuracy metric, which cannot capture the prevalent class overfitting problem in a rare-class data scenario [ 17 , 18 , 19 , 20 , 33 ]. Emerging studies should deliberately consider handling the imbalanced data occurrence and feasibility of building a multi-purpose singular model to handle both brown and white eggs.…”

“…Furthermore, unsupervised classification is always the starting point in any discrimination problem and should be necessarily followed by supervised classification [ 16 ], towards an industrial adoptability consideration. More recent works [ 17 , 18 , 19 , 20 ] have attempted approaches, including the state-of-the-art deep learning architecture, for studying chicken egg fertility. Such works, however, were not successful in fertility detection prior to incubation (day 0 of incubation).…”

An Adaptive Partial Least Square Regression Approach for Classifying Chicken Egg Fertility by Hyperspectral Imaging

“…These results showed that the PLS algorithm used can discriminate both brown and white fertile eggs from non-fertile eggs prior to incubation, using any of the thresholds identified. This is a huge contribution when compared to earlier studies that have not achieved such accuracies prior to incubation, with some studies reporting only the overall accuracy metric, which cannot capture the prevalent class overfitting problem in a rareclass data scenario [17][18][19][20]33]. Emerging studies should deliberately consider handling the imbalanced data occurrence and feasibility of building a multi-purpose singular model to handle both brown and white eggs.…”

“…Furthermore, unsupervised classification is always the starting point in any discrimination problem and should be necessarily followed by supervised classification [16], towards an industrial adoptability consideration. More recent works [17][18][19][20] have attempted approaches, including the state-of-the-art deep learning architecture, for studying chicken egg fertility. Such works, however, were not successful in fertility detection prior to incubation (day 0 of incubation).…”

Non-Destructive Assessment of Chicken Egg Fertility

“…Likewise, it was not stated clearly how the different classes of fertile, infertile, and dead embryo eggs was determined. In the same vein, a follow up study to [ 27 ] by [ 28 ] adopted a dense pixelwise spatial attention (DPSA) network for hatching egg fertility classification. The same methodology as in [ 27 ] was followed and so the questions regarding image data source and classes determination remain unanswered.…”