For the deployment of Convolutional Neural Networks (CNNs) on battery-powered, energy-constrained edge devices, both weights and activations in a network can be quantized to reduce the energy consumption associated with CNN inference, as low-precision integer arithmetic uses less energy to execute than operations on full-precision floating-point data. However, this incurs a performance/accuracy degradation. Recently, some quantization algorithms have been introduced which are closing the performance gap with full precision neural networks.
In this work, the student will implement one of these methods and evaluate their performance on a selected benchmark application. Furthermore, a comparison of your results will be drawn with our currently implemented methods.
- Machine Learning, experience with Pytorch
- Proficiency in Python
- Basics of Computer Vision