Lightweight surveillance systems on embedded domains
This report aims to investigate the development of a human detection machine vision algorithm that is lightweight and optimized for resource-constrained embedded systems. Pre-trained lightweight models as a base to implement real-time human detection on the Himax WE-I plus board as the testbed. The...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/181568 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This report aims to investigate the development of a human detection machine vision algorithm that is lightweight and optimized for resource-constrained embedded systems. Pre-trained lightweight models as a base to implement real-time human detection on the Himax WE-I plus board as the testbed. The project's main goal is to have a model, which is very lightweight and uses the necessary optimization techniques like quantization and pruning, that has a small model size and has good speed and performance without sacrificing the model's accuracy. Different platforms that can simplify the process of creating the lightweight model were explored. The findings of the project demonstrate that with appropriate optimizations, deep learning models can be implemented on resource-constrained embedded systems which enables applications such as smart surveillance systems to function effectively in real-world environments. |
|---|