Abstract
Analysis of locomotion is often used as a measure for impairment and recovery following experimental peripheral nerve injury. Compared to rodents, sheep offer several advantages for studying peripheral nerve regeneration. In the present study, we compared for the first time, two-dimensional (2D) and three-dimensional (3D) hindlimb kinematics during obstacle avoidance in the ovine model. This study obtained kinematic data to serve as a template for an objective assessment of the ankle joint motion in future studies of common peroneal nerve (CP) injury and repair in the ovine model. The strategy used by the sheep to bring the hindlimb over a moderately high obstacle, set to 10% of its hindlimb length, was pronounced knee, ankle and metatarsophalangeal flexion when approaching and clearing the obstacle. Despite the overall time course kinematic patterns about the hip, knee, ankle, and metatarsophalangeal were identical, we found significant differences between values of the 2D and 3D joint angular motion. Our results showed that the most apparent changes that occurred during the gait cycle were for the ankle (2D-measured STANCEmax: 157 +/- 2.4 degrees vs. 3D-measured STANCEmax: 151 +/- 1.2 degrees; P<.05) and metatarsophalangeal joints (2D-measured STANCEmin: 151 +/- 2.2 degrees vs. 3D-measured STANCEmin: 162 +/- 2.2 degrees; P<.01 and 2D-measured TO: 163 +/- 4.9 degrees vs. 3D-measured TO: 177 +/- 1.4 degrees; P<.05), whereas the hip and knee joints were much less affected. Data and techniques described here are useful for an objective assessment of altered gait after CP injury and repairin an ovine model.

Abstract
Mobile robotics in forests is currently a hugely important topic due to the recurring appearance of forest wildfires. Thus, in-site management of forest inventory and biomass is required. To tackle this issue, this work presents a study on detection at the ground level of forest tree trunks in visible and thermal images using deep learning-based object detection methods. For this purpose, a forestry dataset composed of 2895 images was built and made publicly available. Using this dataset, five models were trained and benchmarked to detect the tree trunks. The selected models were SSD MobileNetV2, SSD Inception-v2, SSD ResNet50, SSDLite MobileDet and YOLOv4 Tiny. Promising results were obtained; for instance, YOLOv4 Tiny was the best model that achieved the highest AP (90%) and F1 score (89%). The inference time was also evaluated, for these models, on CPU and GPU. The results showed that YOLOv4 Tiny was the fastest detector running on GPU (8 ms). This work will enhance the development of vision perception systems for smarter forestry robots.

Abstract
One of the principal complications of patients that suffer from Diabetes Mellitus (DM) and that can lead to ulceration is the Diabetic foot. As tissue inflammation causes temperature variation, several studies show that thermography can be used to detect complications in diabetic foot and help predicting the risk of ulceration. It is known that, although healthy individuals present characteristic plantar temperature variation patterns, the same does not happen with diabetic patients, for which a particular pattern can not be found; thus, making the measurement of the temperature variation more difficult. Given that, it is important to research in this field in order to obtain methods that can detect atypical variations of the temperature in the sole of the foot. With this in mind, the objective of this work is to present a methodology to analyze the distribution of temperature in thermograms of the foot's plant and classify it as belonging to a DM individual with risk of ulceration or a healthy individual. After foot partitioning with a clustering algorithm, basic statistical descriptors are computed for each cluster. A binary classifier to predict the risk of ulceration in the diabetic foot was evaluated with the different descriptors; both a quantitative temperature index and a classification threshold are calculated for each descriptor. To evaluate the performance of the classifier, experiments were conducted using a public dataset (containing 45 thermograms of healthy individuals and 122 images of DM ones); the following metrics were obtained: Accuracy = 78%, AUC = 86% and F-measure = 84%, with the best descriptor.

Abstract
Robotics navigation and perception for forest management are challenging due to the existence of many obstacles to detect and avoid and the sharp illumination changes. Advanced perception systems are needed because they can enable the development of robotic and machinery solutions to accomplish a smarter, more precise, and sustainable forestry. This article presents a state-of-the-art review about unimodal and multimodal perception in forests, detailing the current developed work about perception using a single type of sensors (unimodal) and by combining data from different kinds of sensors (multimodal). This work also makes a comparison between existing perception datasets in the literature and presents a new multimodal dataset, composed by images and laser scanning data, as a contribution for this research field. Lastly, a critical analysis of the works collected is conducted by identifying strengths and research trends in this domain.

Abstract
Convolutional neural networks (CNN) and transfer learning are receiving a lot of attention because of the positive results achieved on image recognition and classification. Hip dysplasia is the most prevalent hereditary orthopedic disease in the dog. The definitive diagnosis is using the hip radiographic image. This article compares the results of the conventional canine hip dysplasia (CHD) classification by a radiologist using the Fédération Cynologique Internationale criteria and the computer image classification using the Inception-V3, Google’s pre-trained CNN, combined with the transfer learning technique. The experiment’s goal was to measure the accuracy of the model on classifying normal and abnormal images, using a small dataset to train the model. The results were satisfactory considering that, the developed model classified 75% of the analyzed images correctly. However, some improvements are desired and could be achieved in future works by developing a software to select areas of interest from the hip joints and evaluating each hip individually.

Abstract