Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/107806
Title: Detección y clasificación de células normales de la sangre periférica usando aprendizaje profundo
Author: Castro Zapata, Wilson Alfredo
Tutor: Alférez, Santiago  
Abstract: Through the use of digital image processing tools, quantitative descriptors of cells in peripheral blood can be extracted, which, through Machine Learning (M.L.) and Deep Learning (D.L.) techniques, allow cell types to be automatically and objectively differentiated. In this TFM, a system of detection and classification of normal blood cells is trained through convolutional neural networks (CNN) which, when an image is entered, detects the main cells and then allows the leukocytes can be trimmed and classified according to the type. To achieve greater efficiency and overcome the difficulties that arise in the tasks of detection and classification, especially in the determination of descriptors that are made by hand, the problem is addressed with the use of convolutional neural network architectures with the which these descriptors are obtained automatically. For the development of the classifier a network was trained from scratch with an accuracy of 0.862 and a val_acuraccy of 0.858. Other models were reentrained with networks such as VGG16 and Mobile and the results were compared. The latest versions of Keras architecture with Tensorflow as backend was used in both trainings. For the training of the detector, the YOLO architecture was used with the specialized ImageAI library in object detection and excellent performance has been obtained. These developments have various applications in the fields of bioinformatics, and might provide the basis for the application of methodologies similar to those proposed in this TFM, so CNN could be trained to identify pathologies from the classification and detection of abnormal blood cells.
Keywords: image classification
object detection
deep learning
Document type: info:eu-repo/semantics/masterThesis
Issue Date: 8-Jan-2020
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Trabajos finales de carrera, trabajos de investigación, etc.

Files in This Item:
File Description SizeFormat 

PresentaciónTFM_WCZ.mp4


  Restricted Access
443,42 MBMP4View/Open Request a copy
wcastrozTFM0120memoria.pdfMemoria del TFM3,39 MBAdobe PDFThumbnail
View/Open