Carolina Cuesta-Lázaro
Arnau Quera-Bofarull
(Joseph Bullock)
2 months team project at IBEX innovations
Carolina / Arnau
Cosmology
Joe
Particle Physics
Detect
collimator
Segment
Open beam
Bone
Soft-tissue
Kazeminia, S., Karimi, et al (2015)
Extracts features from a high dimensional feature space, once trained on a particular dataset.
DEEP LEARNING
Luminosity
Size
Colour
Galaxy
Star
x 2
LOSS = GENERATED OUTPUT - ACTUAL OUTPUT
Galaxy
Star
x 4
Credit : https://www.pnas.org/content/116/4/1074
With wrong data ....
150 labeled images.
Hardware limitations (memory, training time...).
We need a fast network, easier to re-train as we get more images.
CONS
PROS
Could work for different detectors (different noise).
Generalize to different body-parts.
Well defined boundaries between regions.
Could be improved through more training.
Coursera
Solution:
Artificially augment the dataset by transforming original images.
We typically divide our dataset into three subsets:
First attempts focused on a very simplified SegNet model.
Going deeper has limits ( limited image size, GPU memory bottleneck, overfitting).
Ways to reduce overfitting
Idea:
Penalise the network if it uses too many parameters to fit the data.
Credit: www.kdnuggets.com
The network outputs 3 probability maps.
Soft tissue probability map
We can reduce the number of false positives by making a probability cut to the map.
Probability
Cryptocurrency Times
Paper out arXiv:1812.00548v1, and presented at the SPIE Medical Imaging conference in San Diego. Best student paper awarded.