[Workshop] tinyML Talks: Low-Power Computer Vision
tinyML Talks - Yung-Hsiang Lu: Low-Power Computer Vision1
Hierarchical Neural Networks
By utilizing hierarchical neural network, we can separate the big neural network into much small ones, hence reduce the training time and inference power consumption. However, it might increase the latency.
The Existing Technologies for Making Computer Vision Energy-Efficient
This presentation also surveyed the existing technologies for making computer vision energy-efficient, including
- parameter quantization and pruning;
- compressed convolutional filters and matrix factorization;
- network architecture search;
- knowledge distillation.