Recently, many Deep Learning models have been employed to classify different kinds of plant diseases, but very little work has been done for disease severity detection. However, it is more important to master the severities of plant diseases accurately and timely, as it helps to make effective decisions to protect the plants from being further infected and reduce financial loss. In this paper, based on the Huanglongbing (HLB)-infected leaf images obtained from PlantVillage and crowdAI, we created a dataset with 5,406 citrus leaf images infected by HLB. Then six popular models were trained to perform the severity detection of citrus HLB with the goal to find which types of models are more suitable to detect HLB severity with the same training circumstance. The experimental results show that the Inception_v3 model with epochs=60 can achieve higher accuracy than that of other models for severity detection with an accuracy of 74.38% due to its highly computational efficiency and small number of parameters. Additionally, aiming for evaluating whether the augmented data can contribute to improve the model learning performance, we adopted Generative Adversarial Networks (GANs) to augment the original training dataset up to two times itself. Finally, a new training dataset with 14,056 leaf images composed by the original training images and the augmented ones were used to train the Inception_v3. As a result, we achieved an accuracy of 92.60%, about 20% higher than that of model trained by the original training dataset, which suggested that the GANs-based data augmentation is very useful to improve the model learning performance.

Cephalometric analysis is used to evaluate facial growth, to study the anatomical relationships within the face. Cephalometric assessment is based on 2D radiographic images, either the sagittal or coronal planes and is an inherently inaccurate methodology. The wide availability of 3D imaging techniques, such as computed tomography and magnetic resonance imaging make routine 3D analysis of facial morphology feasible. 3D cephalometry may not only provide a more accurate quantification of the craniofacial morphology and longitudinal growth, but also the differentiation of subtle changes in occlusion. However, a reliable protocol for the computation of craniofacial symmetry and quantification of craniofacial morphology is still a topic of extensive research. Here, a protocol for 3D cephalometric analysis for both the identification of the natural head position (NHP) and the accurate quantification of facial growth and facial asymmetry is proposed and evaluated. A phantom study was conducted to assess the performance of the protocol and to quantify the ability to repeatedly and reliably align skulls with the NHP and quantify the degree of accuracy with which facial growth and facial asymmetry can be measured. The results obtained show that the protocol allows consistent alignment with the NHP, with an overall average error (and standard deviation) of just 0.17 (9.10e-6) mm, with variations of 0.21 (2.77e-17) mm in the frontonasal suture and 0.30 (5.55e-17) mm in the most prominent point in the chin. The average errors associated with simulated facial growth ranged from 1.83 to 3.75% for 2 years’ growth and from − 9.57 to 14.69% for 4 years, while the error in the quantification of facial asymmetry ranged from − 11.38 to 9.31%. The protocol for 3D skull alignment produces accurate and landmark free estimation of the true symmetry of the head. It allows a reliable alignment of the skull in the NHP independently of user-defined landmarks, as well as an accurate quantification of facial growth and asymmetry.

Surface-based geometric modeling has many advantages in terms of visualization and traditional subtractive manufacturing using computer-numerical-control cutting-machine tools. However, it is not an ideal solution for additive manufacturing because to digitally print a surface-represented geometric object using a certain additive manufacturing technology, the object has to be converted into a solid representation. However, converting a known surface-based geometric representation into a printable representation is essentially a redesign process, and this is especially the case, when its interior material structure needs to be considered. To specify a 3D geometric object that is ready to be digitally manufactured, its representation has to be in a certain volumetric form. In this research, we show how some of the difficulties experienced in additive manufacturing can be easily solved by using implicitly represented geometric objects. Like surface-based geometric representation is subtractive manufacturing-friendly, implicitly described geometric objects are additive manufacturing-friendly: implicit shapes are 3D printing ready. The implicit geometric representation allows to combine a geometric shape, material colors, an interior material structure, and other required attributes in one single description as a set of implicit functions, and no conversion is needed. In addition, as implicit objects are typically specified procedurally, very little data is used in their specifications, which makes them particularly useful for design and visualization with modern cloud-based mobile devices, which usually do not have very big storage spaces. Finally, implicit modeling is a design procedure that is parallel computing-friendly, as the design of a complex geometric object can be divided into a set of simple shape-designing tasks, owing to the availability of shape-preserving implicit blending operations.