Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/selimchraibi/real-time-human-detection
Learning localisation without localisation data
https://github.com/selimchraibi/real-time-human-detection
Last synced: 13 days ago
JSON representation
Learning localisation without localisation data
- Host: GitHub
- URL: https://github.com/selimchraibi/real-time-human-detection
- Owner: SelimChraibi
- Created: 2018-04-28T14:09:12.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2023-10-04T14:43:44.000Z (about 1 year ago)
- Last Synced: 2023-10-05T01:56:58.713Z (about 1 year ago)
- Language: Jupyter Notebook
- Homepage:
- Size: 109 MB
- Stars: 3
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Learning localisation without localisation data
*Based on Zhou et al, 2015 [Learning Deep Features for Discriminative Localization](https://arxiv.org/abs/1512.04150).*
Neural Networks are often described as black boxes. This project however, presents a method based on the interpretation of the internal parameters of a neural network to implement an application capable of:
- Detecting the presence of humans in a live video
- Identifying *"regions of interest"* where the individuals detected are most likely to be situated
This method was introduced in the following paper: [Learning Deep Features for Discriminative Localization](http://cnnlocalization.csail.mit.edu/Zhou_Learning_Deep_Features_CVPR_2016_paper.pdf).
Its originality comes from the simplicity of both the network it uses and of the training it requires.
- An explanation of this method and of the specific problem it is applied to in this project can be found in the [`Project Report`](https://SelimChraibi.github.io/real-time-human-detection/) page.
[](https://SelimChraibi.github.io/real-time-human-detection/)
- A demonstration of how the code shared in this repo can be used (to create classification models capable of outputting localisation information) can be found in the jupyter notebook [`demo.ipynb`](https://nbviewer.jupyter.org/github/SelimChraibi/real-time-human-detection/blob/master/demo.ipynb).
[](https://nbviewer.jupyter.org/github/SelimChraibi/real-time-human-detection/blob/master/demo.ipynb)
## Getting started with the code
Create a conda environment from the `yml` file and activate it:
``` bash
conda env create -f environment.yml
conda activate ml-environement
```
You should be ready to use the `webcam_cam.py` app and the `demo.ipynb` notebook.
> Note: CUDA and cuDNN are required for GPU support and don’t come with the conda environment mentioned above.
>- [Installing CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html)
>- [Installing cuDNN](https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html)
## Application using the trained models
Launching the live webcam application:
``` sh
python3 webcam_cam.py --model ./saved_model/mobilenet_with_gi_data.h5
```
>**Requires:**
>Tensorflow version >= 1.7
>Keras version >= 2.1
>OpenCV version >= 3.3