IOT elderly fall detector
The purpose of this report investigates and implement different state-of-the-art deep learning network models for the elderly fall detector CCTV. We will be using 2 models to detect the humans pose estimation in the CCTV video. The first model that we will be using is a pretrain faster RCNN model w...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/162670 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | The purpose of this report investigates and implement different state-of-the-art deep learning network models for the elderly fall detector CCTV.
We will be using 2 models to detect the humans pose estimation in the CCTV video. The first model that we will be using is a pretrain faster RCNN model which is a deep convolutional neural network used for object detection. This model is developed by a group of Microsoft Research. The faster RCNN model can identify the locations of different objects precisely and quickly. We will be using this model to detect the human object in the CCTV camera.
The second model that we will be using is Resnet 50 which was develop in 2015 by Kaiming He et al from MS research team. This model won the ImageNet competition in 2015. This model proves that very deep network layer can work too. In this project, we will use the transfer learning method with the backbone of pretrain Resnet50 model to train the data and use heatmap to predict the pose estimation of human joints using the COCO dataset annotation of 17 human joints key point. |
|---|