In this thesis, we present ContextNet, a novel general object detection framework for incorporating context cues into a detection pipeline. Current deep learning methods for object detection exploit state-of-the-art image recognition networks for classifying the given region-of-interest (ROI) to predefined classes and regressing a bounding-box around it without using any information about the corresponding scene. ContextNet is based on an intuitive idea of having cues about the general scene (e.g., kitchen and library), and changes the priors about presence/absence of some object classes. We provide a general means for integrating this notion in the decision process about the given ROI by using a pretrained network on the scene recognition datasets in parallel to a pretrained network for extracting object-level features for the corresponding ROI. Using comprehensive experiments on the PASCAL VOC 2007, we demonstrate the effectiveness of our design choices, the resulting system outperforms the baseline in most object classes, and reaches 57.5 mAP (mean Average Precision) on the PASCAL VOC 2007 test set in comparison with 55.6 mAP for the baseline. / MS

Neglected Tropical Diseases (NTDs) currently affect approximately 1.6 billion people worldwide, predominantly impacting populations with limited resources and access to healthcare. The study employs an interdisciplinary approach within the field of diagnostics and information technology to investigate the application of computer vision in developing diagnostic tools with the aim of fighting the spread of neglected tropical diseases (NTDs). By leveraging advancements in the field of computer vision, the research seeks to enhance diagnostic accuracy and efficiency by lowering the usage barriers of the diagnostic tool. The research explores the feasibility of using computer vision to differentiate between various characteristics of images generated by a microscope in a diagnostic setting. The aim is to determine the most suitable method for image analysis in the diagnostic setting, comparing conventional image processing techniques, such as image filtering and color models, with Artificial Intelligence (AI)-based methods. The results revealed that the complexity of the images rendered conventional image filters and color models inadequate, highlighting the necessity of alternative methodologies, such as AI. The findings suggest that AI-based approaches are better suited to handle the intricate details and variations present in the images captured by the microscope by offering more accurate and reliable diagnostic capabilities. However, the model trained on single-labeled images required an additional technique for addressing images containing multiple characteristics, namely thresholding. Thresholds were essential for effecting the model's final prediction to suit the specific use case. By implementing thresholds, the model could, to a higher degree, distinguish between overlapping features within the images, ensuring more accurate classification and enhancing overall performance in the diagnostic setting. The final result presents a promising AI model that has the potential to reduce the usage barriers of the diagnostic tool. Hence, this study represents a small step in the right direction toward the larger goal of fighting the spread of neglected tropical diseases.

Modern medical imaging techniques have revolutionized health care over the last decades, providing clinicians with high-resolution 3D images of the inside of the patient's body without the need for invasive procedures. Detailed images of the vascular anatomy can be captured by angiography, providing a valuable source of information when deciding whether a vascular intervention is needed, for planning treatment, and for analyzing the success of therapy. However, increasing level of detail in the images, together with a wide availability of imaging devices, lead to an urgent need for automated techniques for image segmentation and analysis in order to assist the clinicians in performing a fast and accurate examination. To reduce the need for user interaction and increase the speed of vascular segmentation,  we propose a fast and fully automatic vascular skeleton extraction algorithm. This algorithm first analyzes the volume's intensity histogram in order to automatically adapt the internal parameters to each patient and then it produces an approximate skeleton of the patient's vasculature. The skeleton can serve as a seed region for subsequent surface extraction algorithms. Further improvements of the skeleton extraction algorithm include the expansion to detect the skeleton of diseased arteries and the design of a convolutional neural network classifier that reduces false positive detections of vascular cross-sections. In addition to the complete skeleton extraction algorithm, the thesis presents a segmentation algorithm based on modified onion-kernel region growing. It initiates the growing from the previously extracted skeleton and provides a rapid binary segmentation of tubular structures. To provide the possibility of extracting precise measurements from this segmentation we introduce a method for obtaining a segmentation with subpixel precision out of the binary segmentation and the original image. This method is especially suited for thin and elongated structures, such as vessels, since it does not shrink the long protrusions. The method supports both 2D and 3D image data. The methods were validated on real computed tomography datasets and are primarily intended for applications in vascular segmentation, however, they are robust enough to work with other anatomical tree structures after adequate parameter adjustment, which was demonstrated on an airway-tree segmentation.

Submitted by PPG Ci?ncia da Computa??o (ppgcc@pucrs.br) on 2018-06-15T16:40:47Z No. of bitstreams: 1 LEANDRO PEREIRA DA SILVA_DIS.pdf: 16008947 bytes, checksum: 327a925ea56fcca0a86530a0eb3b1637 (MD5) / Approved for entry into archive by Sheila Dias (sheila.dias@pucrs.br) on 2018-06-26T13:25:28Z (GMT) No. of bitstreams: 1 LEANDRO PEREIRA DA SILVA_DIS.pdf: 16008947 bytes, checksum: 327a925ea56fcca0a86530a0eb3b1637 (MD5) / Made available in DSpace on 2018-06-26T13:34:22Z (GMT). No. of bitstreams: 1 LEANDRO PEREIRA DA SILVA_DIS.pdf: 16008947 bytes, checksum: 327a925ea56fcca0a86530a0eb3b1637 (MD5) Previous issue date: 2018-03-27 / Computer vision is the science that aims to give computers the capability of see- ing the world around them. Among its tasks, object recognition intends to classify objects and to identify where each object is in a given image. As objects tend to occur in particular environments, their contextual association can be useful to improve the object recognition task. To address the contextual awareness on object recognition task, the proposed ap- proach performs the identification of the scene context separately from the identification of the object, fusing both information in order to improve the object detection. In order to do so, we propose a novel architecture composed of two convolutional neural networks running in parallel: one for object identification and the other to the identification of the context where the object is located. Finally, the information of the two-streams architecture is concatenated to perform the object classification. The evaluation is performed using PASCAL VOC 2007 and MS COCO public datasets, by comparing the performance of our proposed approach with architectures that do not use the scene context to perform the classification of the ob- jects. Results show that our approach is able to raise in-context object scores, and reduces out-of-context objects scores. / A vis?o computacional ? a ci?ncia que permite fornecer aos computadores a ca- pacidade de verem o mundo em sua volta. Entre as tarefas, o reconhecimento de objetos pretende classificar objetos e identificar a posi??o onde cada objeto est? em uma imagem. Como objetos costumam ocorrer em ambientes particulares, a utiliza??o de seus contex- tos pode ser vantajosa para melhorar a tarefa de reconhecimento de objetos. Para utilizar o contexto na tarefa de reconhecimento de objetos, a abordagem proposta realiza a iden- tifica??o do contexto da cena separadamente da identifica??o do objeto, fundindo ambas informa??es para a melhora da detec??o do objeto. Para tanto, propomos uma nova arquite- tura composta de duas redes neurais convolucionais em paralelo: uma para a identifica??o do objeto e outra para a identifica??o do contexto no qual o objeto est? inserido. Por fim, a informa??o de ambas as redes ? concatenada para realizar a classifica??o do objeto. Ava- liamos a arquitetura proposta com os datasets p?blicos PASCAL VOC 2007 e o MS COCO, comparando o desempenho da abordagem proposta com abordagens que n?o utilizam o contexto. Os resultados mostram que nossa abordagem ? capaz de aumentar a probabili- dade de classifica??o para objetos que est?o em contexto e reduzir para objetos que est?o fora de contexto.

Autonomous driving is increasingly popular among people and automotive industries in realizing their presence both in passenger and goods transportation. Safer autonomous navigation might be very challenging if there is a failure in sensing system. Among several sensing systems, image classification plays a major role in understanding the road signs and to regulate the vehicle control based on urban road rules. Hence, a robust classifier algorithm irrespective of camera position, view angles, environmental condition, different vehicle size & type (Car, Bus, Truck, etc.,) of an autonomous platform is of prime importance. In this study, Convolutional Neural Network (CNN) based classifier algorithm has been implemented to ensure improved robustness for recognizing traffic signs. As training data play a crucial role in supervised learning algorithms, there come an effective dataset requirement which can handle dynamic environmental conditions and other variations caused due to the vehicle motion (will be referred as challenges). Since the collected training data might not contain all the dynamic variations, the model weakness can be identified by exposing it to variations (Blur, Darkness, Shadow, etc.,) faced by the vehicles in real-time as a initial testing sequence. To overcome the weakness caused due to the training data itself, an effective augmentation technique enriching the training data in order to increase the model capacity for withstanding the variations prevalent in urban environment has been proposed. As a major contribution, a framework has been developed to identify model weakness and successively introduce a targeted augmentation methodology for classification improvement. Targeted augmentation is based on estimated weakness caused due to the challenges with difficulty levels, only those necessary for better classification were then augmented further. Predictive Augmentation (PA) and Predictive Multiple Augmentation (PMA) are the two proposed methods to adapt the model based on targeted challenges by delivering with high numerical value of confidence. We validated our framework on two different training datasets (German Traffic Sign Recognition Benchmark (GTSRB) and Heavy Vehicle data collected from bus) and with 5 generated test groups containing varying levels of challenge (simple to extreme). The results show impressive improvement by ≈ 5-20% in overall classification accuracy thereby keeping their high confidence.

Submitted by Angela Maria de Oliveira (amolivei@uepg.br) on 2019-02-28T17:58:29Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Renan Rosa.pdf: 4067327 bytes, checksum: eb0bd9e84fbd89a24b4a397c9655fa62 (MD5) / Made available in DSpace on 2019-02-28T17:58:29Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) Renan Rosa.pdf: 4067327 bytes, checksum: eb0bd9e84fbd89a24b4a397c9655fa62 (MD5) Previous issue date: 2018-11-19 / As pragas em lavouras causam prejuízos econômicos na agricultura, reduzindo a produção e consequentemente os lucros. O manejo de pragas é essencial, para reduzir estes prejuízos, e consiste na identificação e posterior controle desse tipo de ameaça. O controle é fundamentalmente dependente da identificação, pois é a partir dela que o manejo é feito. A identificação é feita visualmente, baseando-se nas características da praga. Essas características são inerentes e diferem de espécie para espécie. Devido à dificuldade da identificação, esse processo é realizado principalmente por profissionais especializados na área, o que acarreta na concentração do conhecimento. Esta dissertação apresenta uma metodologia para classificação de pragas por meio de técnicas de computação, onde um sistema computacional do tipo clienteservidor foi criado a fim de prover a classificação de pragas por meio de serviço, que é realizado pelo uso de rede neural convolucional baseada na arquitetura Inception V3. As pragas Anticarsia Gemmatalis, Helicoverpa armigera e Spodoptera Cosmioides, foram escolhidas para classificação por serem bastante comuns no estado do Paraná. A rede neural convolucional obteve índice de acerto de 92,5%. / Pests on crops cause economic damage to agriculture, reducing production and consequently profits. Pest management is essential to reduce these losses, and consists in the identification and subsequent control of this type of threat. Control is fundamentally dependent on identification, because management is done from it. The identification is made visually, based on the characteristics of the pest. These characteristics are inherent and differ from species to species. Due to the difficulty of identification, this process is carried out mainly by professionals specialized in the area, which entails the concentration of knowledge. This dissertation presents a methodology for pest classification by means of computational techniques, in which a client-server computational system was created in order to provide pest classification by means of a service, which is performed by the use of convolutional neural network based in the Inception V3 architecture. The pests Anticarsia Gemmatalis, Helicoverpa armigera and Spodoptera Cosmioides, were chosen for classification because they are quite common in the state of Paraná. The convolutional neural network obtained a success rate of 92.5%.

Submitted by JOSIANE SANTOS DE OLIVEIRA (josianeso) on 2017-06-14T14:08:02Z No. of bitstreams: 1 Uilian Kenedi Lopes_.pdf: 1606525 bytes, checksum: 6cf55b065e939fca4ecdc6564199c594 (MD5) / Made available in DSpace on 2017-06-14T14:08:02Z (GMT). No. of bitstreams: 1 Uilian Kenedi Lopes_.pdf: 1606525 bytes, checksum: 6cf55b065e939fca4ecdc6564199c594 (MD5) Previous issue date: 2017-03-24 / Nenhuma / De acordo com a Organização Mundial de Saúde, a tuberculose (juntamente com a AIDS) é a doença infecciosa que mais causa mortes no mundo. Estima-se que em 2014 cerca de 1,5 milhão de pessoas infectadas com o Mycobacterium Tuberculosis morreram, a maior parte delas nos países em desenvolvimento. Muitas destas mortes poderiam ter sido evitadas caso o diagnóstico ocorresse nas fases iniciais da doença, mas infelizmente as técnicas mais avançadas de diagnóstico ainda têm custo proibitivo para adoção em massa nos países em desenvolvimento. Uma das técnicas mais populares de diagnóstico da tuberculose ainda é através da radiografia torácica frontal, entretanto este método tem seu impacto reduzido devido à necessidade de radiologistas treinados analisarem cada radiografia individualmente. Por outro lado, já existem pesquisas buscando automatizar o diagnóstico através da aplicação de técnicas computacionais às imagens radiográficas pulmonares, eliminando assim a necessidade da análise individual de cada radiografia e diminuindo grandemente o custo. Além disso, aprimoramentos recentes nas Redes Neurais Convolucionais, relacionados também à área de Deep Learning, obtiveram grande sucesso para classificação de imagens nos mais diversos domínios, porém sua aplicação no diagnóstico da tuberculose ainda é limitada. Assim o foco deste trabalho é produzir uma investigação que promova avanços nas pesquisas, trazendo três abordagens de aplicação de Redes Neurais Convolucionais com objetivo de detectar a doença. As três propostas apresentadas neste trabalho são implementadas e comparadas com a literatura corrente. Os resultados obtidos até o momento mostraram-se sempre competitivos com trabalhos já publicados na área, obtendo resultados superiores na maior parte dos casos, demonstrando assim o grande potencial das Redes Convolucionais como extratoras de características de imagens médicas. / According to the World Health Organization, tuberculosis (along with AIDS) is the most deadly infectious disease in the world. In 2014 it is estimated that 1.5 million people infected by the Mycobacterium Tuberculosis died, most of them in developing countries. Many of those deaths could have been prevented if the disease was detected at an earlier stage, but unfortunately the most advanced diagnosis methods are cost prohibitive for mass adoption in developing countries. One of the most popular tuberculosis diagnosis methods still is by analysis of frontal thoracic radiographies, however the impact of this method is diminished by the need for individual analysis of each radiography by properly trained radiologists. On the other hand, there is significant research on automating diagnosis by the application of computational techniques to lung radiographic images, eliminating the need for individual analysis of the radiographies and greatly diminishing the cost. In addition to that, recent improvements on Convolutional Neural Networks, which are related to Deep Learning, accomplished excellent results classifying images on diverse domains, but it’s application for tuberculosis diagnosis still is limited. Thus, the focus of this work is to produce an investigation that will advance the research in the area, proposing three approaches to the application of Convolutional Neural Networks to detect the disease. The three proposals presented in this works are implemented and compared to the current literature. The obtained results are competitive with works published in the area, achieving superior results in most cases, thus demonstrating the great potential of Convolutional Networks as medical image feature extractors.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-05-30T19:36:59Z No. of bitstreams: 1 GiovanniLucca.pdf: 1608375 bytes, checksum: 90ad0a568a12b861d1a2a3467c275a12 (MD5) / Made available in DSpace on 2017-05-30T19:36:59Z (GMT). No. of bitstreams: 1 GiovanniLucca.pdf: 1608375 bytes, checksum: 90ad0a568a12b861d1a2a3467c275a12 (MD5) Previous issue date: 2017-01-31 / CAPES / Lung cancer is the leading cause of cancer death worldwide, which accounts for more than 17% percent of the total cancer related deaths. However, its early detection may help in a sharp drop in this mortality rate. Because of the arduous analysis process, alternatives such as computational tools that use image processing techniques and pattern recognition have been widely developed and explored for the early diagnosis of this disease, providing a second opinion to the specialist and making this process faster. Therefore, this work proposes a methodology for the diagnosis of slice-based lung nodules extracted from computed tomography images using evolutionary convolutional neural networks. Firstly, the nodules are divided into two sub-regions using the Otsu algorithm based on the particle swarm optimization algorithm. Then, the slices of the nodules and the slices of their sub-regions were resized to the 28 x 28 dimension and given as input simultaneously to the networks. The architecture of the model was composed of three convolutional neural networks sharing the same fully connected layer at the end. Being a parameterized model, the genetic algorithm was applied to optimize some parameters, such as the number of filters in the convolution layers and the number of neurons in the hidden layer. The proposed methodology was tested on the Lung Image Database Consortium and the Image Database Resource Initiative, resulting in a sensitivity of 94.66 %, specificity of 95.14 %, accuracy of 94.78 % and area under the ROC curve of 0.949. / O câncer de pulmão é a maior causa de morte por câncer em todo mundo, representando mais de 17% do total de mortes relacionadas com câncer. No entanto, sua detecçãao precoce pode ajudar em uma queda acentuada nesta taxa de mortalidade. Devido ao árduo processo na análise dos exames por imagens, alternativas como sistemas computacionais que utilizam técnicas de processamento de imagens e reconhecimento de padrões têm sido amplamente desenvolvidos e explorados para o diagnóstico precoce desta doen¸ca, provendo uma segunda opinião para o especialista e tornando esse processo mais rápido. Diante disso, este trabalho propõe uma metodologia para o diagnóstico de nódulos pulmonares baseado nas fatias extraídas da tomografia computadorizada usando as redes neurais convolucionais evolutivas. Primeiramente, os nódulos são divididos em duas sub-regiões utilizando o algoritmo de Otsu baseado no algoritmo de otimização por enxame de partículas. Em seguida, as fatias dos nódulos e as fatias das suas sub-regiões foram redimensionadas para a dimensão 28 x 28 e dadas como entrada simultaneamente às redes. A arquitetura do modelo foi composta por três redes neurais convolucionais compartilhando a mesma camada completamente conectada no final. Tratando-se de um modelo parametrizado, o algoritmo genético foi aplicado para otimização de alguns parâmetros, tais como a quantidade de filtros nas camadas de convolução e a quantidade de neurônios na camada oculta. A metodologia proposta foi testada na base de imagens Lung Image Database Consortium e a Image Database Resource Initiative, resultando em uma sensibilidade de 94,66%, especifidade de 95,14%, acurácia de 94,78% e área sob a curva ROC de 0,949.

Curbs and road markings were designed to provide a visual low-level spatial perception of road environments. In this sense, a perception system capable of detecting those road features is of utmost importance for an autonomous vehicle. In vision-based approaches, few works have been developed for curb detection, and most of the advances on road marking detection have aimed lane markings only. Therefore, to detect all these road features, multiple algorithms running simultaneously would be necessary. Alternatively, as the main contribution of this work, it was proposed to employ an architecture of Fully Convolutional Neural Network (FCNN), denominated as 3CSeg-Multinet, to detect curbs and road markings in a single inference. Since there was no labeled dataset available for training and validation, a new one was generated with Brazilian urban scenes, and they were manually labeled. By visually analyzing experimental results, the proposed approach has shown to be effective and robust against most of the clutter present on images, running at around 10 fps in a Graphics Processing Unit (GPU). Moreover, with the intention of granting spatial perception, stereo vision techniques were used to project the detected road features in a point cloud. Finally, as a way to validate the applicability of the proposed perception system on a vehicle, it was also introduced a vision-based metric localization model for the urban scenario. In an experiment, compared to the ground truth, this localization method has revealed consistency on its pose estimations in a map generated by LIDAR. / Guias e sinalizações horizontais foram projetados para fornecer a percepção visual de baixo nível do espaço das vias urbanas. Deste modo, seria de extrema importância para um veículo autônomo ter um sistema de percepção capaz de detectar tais características visuais. Em abordagens baseadas em visão, poucos trabalhos foram desenvolvidos para detecção de guias, e a maioria dos avanços em detecção de sinalizações horizontais foi focada na detecção de faixas apenas. Portanto, para que fosse possível detectar todas essas características visuais, seria necessário executar diversos algoritmos simultaneamente. Alternativamente, como sendo a principal contribuição deste trabalho, foi proposto a adoção de uma Rede Neural Totalmente Convolutiva, denominado 3CSeg-Multinet, para detectar guias e sinalizações horizontais em apenas uma inferência. Como não havia um conjunto de dados rotulados disponível para treinar e validar a rede, foi gerado um novo conjunto com imagens capturadas em ambiente urbano brasileiro, e foi realizado a rotulação manual. Através de uma análise visual dos resultados experimentais obtidos, o método proposto mostrou-se eficaz e robusto contra a maioria dos fatores que causam confusão nas imagens, executando a aproximadamente 10 fps em uma GPU. Ainda, com o intuito de garantir a percepção espacial, foram usados métodos de visão estéreo para projetar as características detectadas em núvem de pontos. Finalmente, foi apresentado também um modelo de localização métrica baseado em visão para validar a aplicabilidade do sistema de percepção proposto em um veículo. Em um experimento, este método de localização revelou-se capaz de manter as estimativas consistentes com a verdadeira pose do veículo em um mapa gerado a partir de um sensor LIDAR.

This thesis investigates the application of index based, region segmentation, and deep learning methods to the sensor fusion of omnidirectional (O-D) Infrared (IR) sensors, Kinnect sensors, and O-D vision sensors to increase the level of intelligent perception for unmanned robotic platforms. The goals of this work is first to provide a more robust calibration approach and improve the calibration of low resolution and noisy IR O-D cameras. Then our goal was to explore the best approach to sensor fusion for vegetation detection. We looked at index based, region segmentation, and deep learning methods and compared them with a goal of significant reduction in false positives while maintaining reasonable vegetation detection. The results are as follows: Direct Spherical Calibration of the IR camera provided a more consistent and robust calibration board capture and resulted in the best overall calibration results with sub-pixel accuracy The best approach for sensor fusion for vegetation detection was the deep learning approach, the three methods are detailed in the following chapters with the results summarized here. Modified Normalized Difference Vegetation Index approach achieved 86.74% recognition and 32.5% false positive, with peaks to 80% Thermal Region Fusion (TRF) achieved a lower recognition rate at 75.16% but reduced false positives to 11.75% (a 64% reduction) Our Deep Learning Fusion Network (DeepFuseNet) results demonstrated that deep learning approach showed the best results with a significant (92%) reduction in false positives when compared to our modified normalized difference vegetation index approach. The recognition was 95.6% with 2% false positive. Current approaches are primarily focused on O-D color vision for localization, mapping, and tracking and do not adequately address the application of these sensors to vegetation detection. We will demonstrate the contradiction between current approaches and our deep sensor fusion (DeepFuseNet) for vegetation detection. The combination of O-D IR and O-D color vision coupled with deep learning for the extraction of vegetation material type, has great potential for robot perception. This thesis will look at two architectures: 1) the application of Autoencoders Feature Extractors feeding a deep Convolution Neural Network (CNN) fusion network (DeepFuseNet), and 2) Bottleneck CNN feature extractors feeding a deep CNN fusion network (DeepFuseNet) for the fusion of O-D IR and O-D visual sensors. We show that the vegetation recognition rate and the number of false detects inherent in the classical indices based spectral decomposition are greatly improved using our DeepFuseNet architecture. We first investigate the calibration of omnidirectional infrared (IR) camera for intelligent perception applications. The low resolution omnidirectional (O-D) IR image edge boundaries are not as sharp as with color vision cameras, and as a result, the standard calibration methods were harder to use and less accurate with the low definition of the omnidirectional IR camera. In order to more fully address omnidirectional IR camera calibration, we propose a new calibration grid center coordinates control point discovery methodology and a Direct Spherical Calibration (DSC) approach for a more robust and accurate method of calibration. DSC addresses the limitations of the existing methods by using the spherical coordinates of the centroid of the calibration board to directly triangulate the location of the camera center and iteratively solve for the camera parameters. We compare DSC to three Baseline visual calibration methodologies and augment them with additional output of the spherical results for comparison. We also look at the optimum number of calibration boards using an evolutionary algorithm and Pareto optimization to find the best method and combination of accuracy, methodology and number of calibration boards. The benefits of DSC are more efficient calibration board geometry selection, and better accuracy than the three Baseline visual calibration methodologies. In the context of vegetation detection, the fusion of omnidirectional (O-D) Infrared (IR) and color vision sensors may increase the level of vegetation perception for unmanned robotic platforms. A literature search found no significant research in our area of interest. The fusion of O-D IR and O-D color vision sensors for the extraction of feature material type has not been adequately addressed. We will look at augmenting indices based spectral decomposition with IR region based spectral decomposition to address the number of false detects inherent in indices based spectral decomposition alone. Our work shows that the fusion of the Normalized Difference Vegetation Index (NDVI) from the O-D color camera fused with the IR thresholded signature region associated with the vegetation region, minimizes the number of false detects seen with NDVI alone. The contribution of this work is the demonstration of two new techniques, Thresholded Region Fusion (TRF) technique for the fusion of O-D IR and O-D Color. We also look at the Kinect vision sensor fused with the O-D IR camera. Our experimental validation demonstrates a 64% reduction in false detects in our method compared to classical indices based detection. We finally compare our DeepFuseNet results with our previous work with Normalized Difference Vegetation index (NDVI) and IR region based spectral fusion. This current work shows that the fusion of the O-D IR and O-D visual streams utilizing our DeepFuseNet deep learning approach out performs the previous NVDI fused with far infrared region segmentation. Our experimental validation demonstrates an 92% reduction in false detects in our method compared to classical indices based detection. This work contributes a new technique for the fusion of O-D vision and O-D IR sensors using two deep CNN feature extractors feeding into a fully connected CNN Network (DeepFuseNet).