Testing and Verification of the Deep Neural Networks Against Sparse Pixel Defects.
Michal SzczepankiewiczKrystian RadlakKarolina SzczepankiewiczAdam PopowiczPawel ZawistowskiPublished in: SAFECOMP Workshops (2022)
Keyphrases
- neural network
- formal verification
- pattern recognition
- test generation
- genetic algorithm
- symbolic execution
- high dimensional
- input image
- back propagation
- sparse representation
- model checking
- software development life cycle
- verification method
- activation function
- fuzzy logic
- neural network model
- artificial neural networks
- sparse matrix
- defect detection
- training process
- neural nets
- multilayer perceptron
- test suite
- intensity values
- recurrent neural networks
- self organizing maps
- pixel level
- test cases
- dense disparity map
- multi layer
- automated visual inspection