The aim of this paper is to evaluate the use of D-CNN (Deep Convolutional
Neural Networks) algorithms to classify pig body conditions in normal or not
normal conditions, with a focus on characteristics that are observed in
sanitary monitoring, and were used six different algorithms to do this task.
The study focused on five pig characteristics, being these caudophagy, ear
hematoma, scratches on the body, redness, and natural stains (brown or black).
The results of the study showed that D-CNN was effective in classifying
deviations in pig body morphologies related to skin characteristics. The
evaluation was conducted by analyzing the performance metrics Precision,
Recall, and F-score, as well as the statistical analyses ANOVA and the
Scott-Knott test. The contribution of this article is characterized by the
proposal of using D-CNN networks for morphological classification in pigs, with
a focus on characteristics identified in sanitary monitoring. Among the best
results, the average Precision metric of 80.6\% to classify caudophagy was
achieved for the InceptionResNetV2 network, indicating the potential use of
this technology for the proposed task. Additionally, a new image database was
created, containing various pig's distinct body characteristics, which can
serve as data for future research.

评估使用 D-CNN 算法对猪体条件进行分类的研究，主要关注卫生监测中观察到的特征，包括尾部咬食、耳血肿、身体擦伤、发红和自然污渍（棕色或黑色）。研究结果表明，D-CNN 对与皮肤特征相关的猪体形态偏差具有有效分类能力，并通过分析 Precision、Recall、F-score、ANOVA 和 Scott-Knott 测试等性能指标进行评估。该研究的贡献在于提出了在猪类形态学分类中使用 D-CNN 网络的方法，并创建了一个新的图像数据库，可用于未来研究。

基于深度学习的猪形态分类，重点关注卫生监测

Using Deep Learning for Morphological Classification in Pigs with a  Focus on Sanitary Monitoring

Research in texture recognition often concentrates on the problem of material
recognition in uncluttered conditions, an assumption rarely met by
applications. In this work we conduct a first study of material and describable
texture at- tributes recognition in clutter, using a new dataset derived from
the OpenSurface texture repository. Motivated by the challenge posed by this
problem, we propose a new texture descriptor, D-CNN, obtained by Fisher Vector
pooling of a Convolutional Neural Network (CNN) filter bank. D-CNN
substantially improves the state-of-the-art in texture, mate- rial and scene
recognition. Our approach achieves 82.3% accuracy on Flickr material dataset
and 81.1% accuracy on MIT indoor scenes, providing absolute gains of more than
10% over existing approaches. D-CNN easily trans- fers across domains without
requiring feature adaptation as for methods that build on the fully-connected
layers of CNNs. Furthermore, D-CNN can seamlessly incorporate multi-scale
information and describe regions of arbitrary shapes and sizes. Our approach is
particularly suited at lo- calizing stuff categories and obtains
state-of-the-art re- sults on MSRC segmentation dataset, as well as promising
results on recognizing materials and surface attributes in clutter on the
OpenSurfaces dataset.

本文介绍了一种基于卷积神经网络（CNN）的新型纹理描述符 D-CNN，并在嘈杂的环境下完成了纹理和材料识别，并实现了 81.1% 至 82.3% 的识别准确率，显著提高了现有方法的性能。