Single-bit-per-weight deep convolutional neural networks without batch-normalization layers for embedded systems.
Mark D. McDonnellHesham MostafaRunchun WangAndré van SchaikPublished in: CoRR (2019)
Keyphrases
- embedded systems
- convolutional neural networks
- low cost
- resource limited
- computing power
- embedded software
- real time systems
- real time image processing
- processing power
- protocol stack
- software systems
- embedded devices
- flash memory
- hardware software
- multiple layers
- real world
- field programmable gate array
- communication technologies
- hw sw
- embedded real time systems
- instruction set
- convolutional network
- consumer electronics
- safety critical
- smart card
- artificial intelligence