An investigation of Recent Deep Learning Techniques Applied to Blood Cell Image Analysis
(2018) In Master's Theses in Mathematical Sciences FMAM05 20181Mathematics (Faculty of Engineering)
- Abstract
- This project has investigated the performances of Capsule Networks in comparison to Convolutional Neural Networks (CNNs) on white blood cell image classification at CellaVision. The Capsule Network models that were investigated are EM Routing Capsule Networks (EMCNs) and Dynamic Routing Capsule Networks (DCNs). The models were compared with regards to convergence rate, speed, and accuracy performance on datasets with varying size and complexity. The results show that DCNs outperform the other models on small datasets with regards to accuracy and convergence rate, whereas the CNNs outperform the other models on bigger datasets with higher complexity. With regards to speed, CNNs outperform the other models on both CPU and GPU, with DCNs being... (More)
- This project has investigated the performances of Capsule Networks in comparison to Convolutional Neural Networks (CNNs) on white blood cell image classification at CellaVision. The Capsule Network models that were investigated are EM Routing Capsule Networks (EMCNs) and Dynamic Routing Capsule Networks (DCNs). The models were compared with regards to convergence rate, speed, and accuracy performance on datasets with varying size and complexity. The results show that DCNs outperform the other models on small datasets with regards to accuracy and convergence rate, whereas the CNNs outperform the other models on bigger datasets with higher complexity. With regards to speed, CNNs outperform the other models on both CPU and GPU, with DCNs being very slow. EMCNs, meanwhile, give an indication of learning spatial concepts better, as they are more invariant to spatial- and noise transformations of the underlying test datasets. (Less)
- Popular Abstract
- Deep learning is a field which recently has seen enormous improvements in areas such as Image Classification, Robotics and Speech Recognition. One participant in this field is CellaVision, which applies a deep learning technique called Convolutional Neural Networks for 80−90% accuracy on Image Classification of blood cells; helping research labs with automating the diagnosis process. Naturally, CellaVision is interested in finding better classification algorithms, which is why this project has researched the recent paradigm of Capsule Networks. The investigations show that these networks learn spatial concepts better and outperform Convolutional Neural Networks on small data sets, but otherwise perform equally well to - or worse than... (More)
- Deep learning is a field which recently has seen enormous improvements in areas such as Image Classification, Robotics and Speech Recognition. One participant in this field is CellaVision, which applies a deep learning technique called Convolutional Neural Networks for 80−90% accuracy on Image Classification of blood cells; helping research labs with automating the diagnosis process. Naturally, CellaVision is interested in finding better classification algorithms, which is why this project has researched the recent paradigm of Capsule Networks. The investigations show that these networks learn spatial concepts better and outperform Convolutional Neural Networks on small data sets, but otherwise perform equally well to - or worse than Convolutional Neural Networks. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8957638
- author
- Forslöw, Andreas LU
- supervisor
-
- Anders Heyden LU
- Mattias Nilsson LU
- organization
- alternative title
- Undersökning av moderna maskininlärningstekniker för bildbaserad blodcellsanalys
- Capsule Networks applied to White Blood Cell Image Analysis
- course
- FMAM05 20181
- year
- 2018
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- deep learning, machine learning, capsule networks, neural networks, Geoffrey Hinton, Sara Sabour, Hinton, Sabour, Frosst, dynamic routing, em routing, cellavision, andreas forslöw, anders heyden, mattias nilsson, image classification, image analysis, blood cells, white blood cells
- publication/series
- Master's Theses in Mathematical Sciences
- report number
- LUTFMA-3361-2018
- ISSN
- 1404-6342
- other publication id
- 2018:E61
- language
- English
- id
- 8957638
- date added to LUP
- 2018-09-05 14:52:07
- date last changed
- 2018-09-05 14:52:07
@misc{8957638,
abstract = {{This project has investigated the performances of Capsule Networks in comparison to Convolutional Neural Networks (CNNs) on white blood cell image classification at CellaVision. The Capsule Network models that were investigated are EM Routing Capsule Networks (EMCNs) and Dynamic Routing Capsule Networks (DCNs). The models were compared with regards to convergence rate, speed, and accuracy performance on datasets with varying size and complexity. The results show that DCNs outperform the other models on small datasets with regards to accuracy and convergence rate, whereas the CNNs outperform the other models on bigger datasets with higher complexity. With regards to speed, CNNs outperform the other models on both CPU and GPU, with DCNs being very slow. EMCNs, meanwhile, give an indication of learning spatial concepts better, as they are more invariant to spatial- and noise transformations of the underlying test datasets.}},
author = {{Forslöw, Andreas}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master's Theses in Mathematical Sciences}},
title = {{An investigation of Recent Deep Learning Techniques Applied to Blood Cell Image Analysis}},
year = {{2018}},
}