Indiana University-Purdue University Indianapolis (IUPUI) / Domain adaptation is one of the hottest directions in solving annotation insufficiency problem of deep learning. General domain adaptation is not consistent with the practical scenarios in the industry. In this thesis, we focus on two concerns as below. First is that labeled data are generally collected from multiple domains. In other words, multi-source adaptation is a more common situation. Simply extending these single-source approaches to the multi-source cases could cause sub-optimal inference, so specialized multi-source adaptation methods are essential. The main challenge in the multi-source scenario is a more complex divergence situation. Not only the divergence between target and each source plays a role, but the divergences among distinct sources matter as well. However, the significance of maintaining consistency among multiple sources didn't gain enough attention in previous work. In this thesis, we propose an Enhanced Consistency Multi-Source Adaptation (EC-MSA) framework to address it from three perspectives. First, we mitigate feature-level discrepancy by cross-domain conditional alignment, narrowing the divergence between each source and target domain class-wisely. Second, we enhance multi-source consistency via dual mix-up, diminishing the disagreements among different sources. Third, we deploy a target distilling mechanism to handle the uncertainty of target prediction, aiming to provide high-quality pseudo-labeled target samples to benefit the previous two aspects. Extensive experiments are conducted on several common benchmark datasets and demonstrate that our model outperforms the state-of-the-art methods. Second is that data privacy and security is necessary in practice. That is, we hope to keep the raw data stored locally while can still obtain a satisfied model. In such a case, the risk of data leakage greatly decreases. Therefore, it is natural for us to combine the federated learning paradigm with domain adaptation. Under the source-private setting, the main challenge for us is to expose information from the source domain to the target domain while make sure that the communication process is safe enough. In this thesis, we propose a method named Fourier Transform-Assisted Federated Domain Adaptation (FTA-FDA) to alleviate the difficulties in two ways. We apply Fast Fourier Transform to the raw data and transfer only the amplitude spectra during the communication. Then frequency space interpolations between these two domains are conducted, minimizing the discrepancies while ensuring the contact of them and keeping raw data safe. What's more, we make prototype alignments by using the model weights together with target features, trying to reduce the discrepancy in the class level. Experiments on Office-31 demonstrate the effectiveness and competitiveness of our approach, and further analyses prove that our algorithm can help protect privacy and security.

Traffic sign recognition systems have been introduced to overcome road-safety concerns. These systems are widely adopted by automotive industry whereby safety critical systems are developed for car manufacturers. To develop an automatic TSDR system is a tedious job given the continuous changes in the environment and lighting conditions. Among the other issues that also need to be addressed are partial obscuring, multiple traffic signs appearing at a single time, and blurring and fading of traffic signs, which can also create problem for the detection purpose . For applying the TSDR system in real-time environment, a fast algorithm is needed. As well as dealing with these issues, a recognition system should also avoid erroneous recognition of no signs. TSDR system would detect and classify a collection of 43 individual traffic-signs taken from real-time environment into different classes for recognition. In this project classification of individual traffic signs is done using deep Convolutional Neural Network with VGG-net architecture model to develop an efficient classifier with improved prediction accuracy (using GTSRB dataset).

No description available.

Indiana University-Purdue University Indianapolis (IUPUI) / The Convolutional Neural Network (CNN) have shown a substantial improvement in the field of Machine Learning. But they do come with their own set of drawbacks. Capsule Networks have addressed the limitations of CNNs and have shown a great improvement by calculating the pose and transformation of the image. Deeper networks are more powerful than shallow networks but at the same time, more difficult to train. Residual Networks ease the training and have shown evidence that they can give good accuracy with considerable depth. Putting the best of Capsule Network and Residual Network together, we present Residual Capsule Network and 3-Level Residual Capsule Network, a framework that uses the best of Residual Networks and Capsule Networks. The conventional Convolutional layer in Capsule Network is replaced by skip connections like the Residual Networks to decrease the complexity of the Baseline Capsule Network and seven ensemble Capsule Network. We trained our models on MNIST and CIFAR-10 datasets and have seen a significant decrease in the number of parameters when compared to the Baseline models.

Indiana University-Purdue University Indianapolis (IUPUI) / Developing intelligent agents that can perceive and understand the rich visualworld around us has been a long-standing goal in the field of AI. Recently, asignificant progress has been made by the CNNs/DNNs to the incredible advances& in a wide range of applications such as ADAS, intelligent cameras surveillance,autonomous systems, drones, & robots. Design space exploration (DSE) of NNs andother techniques have made CNN/DNN memory & computationally efficient. Butthe major design hurdles for deployment are limited resources such as computation,memory, energy efficiency, and power budget. DSE of small DNN architectures forADAS emerged with better and efficient architectures such as baseline SqueezeNetand SqueezeNext. These architectures are exclusively known for their small modelsize, good model speed & model accuracy.In this thesis study, two new DNN architectures are proposed. Before diving intothe proposed architectures, DSE of DNNs explores the methods to improveDNNs/CNNs.Further, understanding the different hyperparameters tuning &experimenting with various optimizers and newly introduced methodologies. First,High Performance SqueezeNext architecture ameliorate the performance of existingDNN architectures. The intuition behind this proposed architecture is to supplantconvolution layers with a more sophisticated block module & to develop a compactand efficient architecture with a competitive accuracy. Second, Shallow SqueezeNextarchitecture is proposed which achieves better model size results in comparison tobaseline SqueezeNet and SqueezeNext is presented. It illustrates the architecture is xviicompact, efficient and flexible in terms of model size and accuracy.Thestate-of-the-art SqueezeNext baseline and SqueezeNext baseline are used as thefoundation to recreate and propose the both DNN architectures in this study. Dueto very small model size with competitive model accuracy and decent model testingspeed it is expected to perform well on the ADAS systems.The proposedarchitectures are trained and tested from scratch on CIFAR-10 [30] & CIFAR-100[34] datasets. All the training and testing results are visualized with live loss andaccuracy graphs by using livelossplot. In the last, both of the proposed DNNarchitectures are deployed on BlueBox2.0 by NXP.

Remotely sensed imagery is one of the most important data sources for large-scale and multi-temporal agricultural, forestry, soil, environmental, social and economic applications. In order to accurately extract useful thematic information of the earth surface from images, various techniques and methods have been developed. The methods can be divided into parametric and non-parametric based on the requirement of data distribution, or into global and local based on the characteristics of modeling global trends and local variability, or into unsupervised and supervised based on whether training data are required, and into design-based and model-based in terms of the theory based on which the estimators are developed. The methods have their own disadvantages that impede the improvement of estimation accuracy. Thus, developing novel methods and improving the existing methods are needed. This dissertation focused on the development of a feature-space indicator simulation (FSIS), the improvement of geographically weighted sigmoidal simulation (GWSS) and k-nearest neighbors (kNN), and their assessment of land use and land cover (LULC) classification and probability (fraction) mapping of percentage vegetation cover (PVC) in Duolun County, Xilingol League, Inner Mongolia, China. The FSIS employs an indicator simulation in a high-dimensional feature space and expends derivation of indicator variograms from geographic space to feature space that leads to feature space indicator variograms (FSIV), to circumvent the issues existing in traditional indicator simulation in geostatistics. The GWSS is a stochastic and probability mapping method and considers a spatially nonstationary sample data and the local variation of an interest variable. The improved kNN, called Optimal k-nearest neighbors (OkNN), searches for an optimal number of nearest neighbors at each location based on local variability, and can be used for both classification and probability mapping. Three methods were validated and compared with several widely used approaches for LULC classification and PVC mapping in the study area. The datasets used in the study included a Landsat 8 image and a total of 920 field plots. The results obtained showed that 1) Compared with maximum likelihood classification (ML), support vector machine (SVM) and random forest (RF), the proposed FSIS classifier led to statistically significantly higher classification accuracy of six LULC types (water, agricultural land, grassland, bare soil, built-up area, and forested area); 2) Compared with linear regression (LR), polynomial regression (PR), sigmoidal regression (SR), geographically weighted regression (GWR), and geographically weighted polynomial regression (GWPR), GWSS did not only resulted in more accurate estimates of PVC, but also greatly reduced the underestimations and overestimation of PVC for the small and large values respectively; 3) Most of the red and near infrared bands relevant vegetation indices had significant contributions to improving the accuracy of mapping PVC; 4) OkNN resulted in spatially variable and optimized k values and higher prediction accuracy of PVC than the global methods; and 5) The range parameter of FSIVs was the major factor that spatially affected the classification accuracy of LULC types, but the FSIVs were less sensitive to the number of training samples. Thus, the results answered all six research questions proposed.

In this dissertation, novel techniques for hyperspectral classification and signal reconstruction from random projections are presented. A classification paradigm designed to exploit the rich statistical structure of hyperspectral data is proposed. The proposed framework employs the local Fisher’s discriminant analysis to reduce the dimensionality of the data while preserving its multimodal structure, followed by a subsequent Gaussianmixture- model or support-vector-machine classifier. An extension of this framework in a kernel induced space is also studied. This classification approach employs a maximum likelihood classifier and dimensionality reduction based on a kernel local Fisher’s discriminant analysis. The technique imposes an additional constraint on the kernel mapping—it ensures that neighboring points in the input space stay close-by in the projected subspace. In a typical remote sensing flow, the sender needs to invoke an appropriate compression strategy for downlinking signals (e.g., imagery to a base station). Signal acquisition using random projections significantly decreases the sender-side computational cost, while preserving useful information. In this dissertation, a novel class-dependent hyperspectral image reconstruction strategy is also proposed. The proposed method employs statistics pertinent to each class as opposed to the average statistics estimated over the entire dataset, resulting in a more accurate reconstruction from random projections. An integrated spectral-spatial model for signal reconstruction from random projections is also developed. In this approach, spatially homogeneous segments are combined with spectral pixel-wise classification results in the projected subspace. An appropriate reconstruction strategy, such as compressive projection principal component analysis (CPPCA), is employed individually in each category based on this integrated map. The proposed method provides better reconstruction performance as compared to traditional methods and the class-dependent CPPCA approach.

Yes / Corneal images can be acquired using confocal microscopes which provide detailed images of the different layers inside the cornea. Most corneal problems and diseases occur in one or more of the main corneal layers: the epithelium, stroma and endothelium. Consequently, for automatically extracting clinical information associated with corneal diseases, or evaluating the normal cornea, it is important also to be able to automatically recognise these layers easily. Artificial intelligence (AI) approaches can provide improved accuracy over the conventional processing techniques and save a useful amount of time over the manual analysis time required by clinical experts. Artificial neural networks (ANN) and adaptive neuro fuzzy inference systems (ANFIS), are powerful AI techniques, which have the capability to accurately classify the main layers of the cornea. The use of an ANFIS approach to analyse corneal layers is described for the first time in this paper, and statistical features have been also employed in the identification of the corneal abnormality. An ANN approach is then added to form a combined committee machine with improved performance which achieves an accuracy of 100% for some classes in the processed data sets. Three normal data sets of whole corneas, comprising a total of 356 images, and seven abnormal corneal images associated with diseases have been investigated in the proposed system. The resulting system is able to pre-process (quality enhancement, noise removal), classify (whole data sets, not just samples of the images as mentioned in the previous studies), and identify abnormalities in the analysed data sets. The system output is visually mapped and the main corneal layers are displayed. 3D volume visualisation for the processed corneal images as well as for each individual corneal cell is also achieved through this system. Corneal clinicians have verified and approved the clinical usefulness of the developed system especially in terms of underpinning the expertise of ophthalmologists and its applicability in patient care.

The effectiveness of colonoscopy depends on the quality of the inspection of the colon. There was no automated measurement method to evaluate the quality of the inspection. This thesis addresses this issue by investigating an automated post-procedure quality measurement technique and proposing a novel approach automatically deciding a percentage of stool areas in images of digitized colonoscopy video files. It involves the classification of image pixels based on their color features using a new method of planes on RGB (red, green and blue) color space. The limitation of post-procedure quality measurement is that quality measurements are available long after the procedure was done and the patient was released. A better approach is to inform any sub-optimal inspection immediately so that the endoscopist can improve the quality in real-time during the procedure. This thesis also proposes an extension to post-procedure method to detect stool, bite-block, and blood regions in real-time using color features in HSV color space. These three objects play a major role in quality measurements in colonoscopy. The proposed method partitions very large positive examples of each of these objects into a number of groups. These groups are formed by taking intersection of positive examples with a hyper plane. This hyper plane is named as 'positive plane'. 'Convex hulls' are used to model positive planes. Comparisons with traditional classifiers such as K-nearest neighbor (K-NN) and support vector machines (SVM) proves the soundness of the proposed method in terms of accuracy and speed that are critical in the targeted real-time quality measurement system.

Indiana University-Purdue University Indianapolis (IUPUI) / Mutual learning algorithm is a machine learning algorithm where multiple machine learning algorithms learns from different sources and then share their knowledge among themselves so that all the agents can improve their classification and prediction accuracies simultaneously. Mutual learning algorithm can be an efficient mechanism for improving the machine learning and neural network efficiency in a multi-agent system. Usually, in knowledge distillation algorithms, a big network plays the role of a static teacher and passes the data to smaller networks, known as student networks, to improve the efficiency of the latter. In this thesis, it is showed that two small networks can dynamically and interchangeably play the changing roles of teacher and student to share their knowledge and hence, the efficiency of both the networks improve simultaneously. This type of dynamic learning mechanism can be very useful in mobile environment where there is resource constraint for training with big dataset. Data exchange in multi agent, teacher-student network system can lead to efficient learning. The concept and the proposed mutual learning algorithm are demonstrated using convolutional neural networks (CNNs) and Support Vector Machine (SVM) to recognize the pattern recognition problem using MNIST hand-writing dataset. The concept of machine learning is applied in the field of natural language processing (NLP) too. Machines with basic understanding of human language are getting increasingly popular in day-to-day life. Therefore, NLP-enabled machines with memory efficient training can potentially become an indispensable part of our life in near future. A classic problem in the field of NLP is news classification problem where news articles from newspapers are classified by news categories by machine learning algorithms. In this thesis, we show news classification implemented using Naïve Bayes and support vector machine (SVM) algorithm. Then we show two small networks can dynamically play the changing roles of teacher and student to share their knowledge on news classification and hence, the efficiency of both the networks improves simultaneously. The mutual learning algorithm is applied between homogenous algorithms first, i.e., between two Naive Bayes algorithms and two SVM algorithms. Then the mutual learning is demonstrated between heterogenous agents, i.e., between one Naïve Bayes and one SVM agent and the relative efficiency increase between the agents is discussed before and after mutual learning. / 2025-04-04