Pedro Henrique Neves da Silva

Determining behavioral deviations in sheep is challenging through visual observation alone, as sheep are generally stoic animals and may not easily show signs of pain or discomfort. This study aimed to evaluate and classify activity changes in sheep using optical flow algorithms as an alternative approach. Video recordings of 32 sheep housed in 8 pens were analyzed over 4 days, with each day consisting of 4 one-hour video segments. These segments were manually observed at 5-minute intervals to document the time spent laying, standing, and walking. Simultaneously, the same video segments were analyzed using optical flow algorithms to track movement patterns. The calculated optical flow measures, including mean, variance, skewness, and kurtosis, were then correlated with the observed behaviors. These measures demonstrated their potential as effective tools for monitoring behavioral activity changes in sheep, providing an objective method to track behavior and detect deviations that might indicate discomfort, stress, or health issues.

An animal nose identification system could allow efficient monitoring of pets, and it can be used in applications such as identifying animal breeds or possible diseases and/or injuries. For that, a bulldog image dataset was build from French bulldogs. To carry out the validation through cross-validation, 10 folds (K-folds) were created. Afterward, five convolutional neural networks (CNN) were trained with our dataset to identify the nose: Faster R-CNN (Region-based CNN), SABL (Side-Aware Boundary Localization), RetinaNet (ResNet50+FPN), VFNet (VarifocalNet), and ATSS (Adaptive Training Sample Selection). Faster R-CNN, SABL, RetinaNet, VFNet and ATSS were used for training in the first phase, while ATSS, Faster R-CNN, ATSS and SABL in the second. The results showed that the ATSS network obtained the highest values of mAP and Accuracy in the first phase. Moreover, SABL network achieved the highest values of mAP50, mAP75, Recall, F-Score and Accuracy at the end of the second phase.

Machine learning (ML) algorithms can be used to predict wood volume in a faster and more accurate way, providing reliable answers in forest inventories. The objective of this work was to evaluate the performance of different ML techniques to predict the volume of eucalyptus wood, using diameter at breast height (DBH) and total height (Ht) as input variables, obtained by measuring DBH and Ht of 72 trees of six eucalyptus species (Eucalyptus camaldulensis, E. uroplylla, E. saligna, E. grandis, E. urograndis, and Corymbria citriodora). The trees were cut down in two different epochs, rendering 48 samples at 24 months and 24 samples at 48 months, and the volume of each tree was measured using the Smailian method. This research explores five machine learning models, namely artificial neural networks (ANN), K-nearest neighbor (KNN), multiple linear regression (LR), random forest (RF) and support vector machine (SVM), to estimate the volume of eucalyptus wood using DBH and Ht. Artificial neural networks achieved higher correlations between observed and estimated wood volume values. However, the RF outperformed all models by providing lower MAE and higher correlations between observed and estimated wood volume values. Therefore, RF is the most accurate for predicting wood volume in eucalyptus species.