Neural network control for active cameras using master-slave setup

Neural network control for active cameras using master-slave setup

The source code associated with the paper “Neural network control for active cameras using master-slave setup“, published in AVSS 2018. The package has the code for a learning-based approach to control the master-slave setup and a framework to compare different methods for master-slave camera system.

Prerequisites:
Python 3.6
Numpy 1.4
Opencv 3.4
Keras 2.1.5
Tensorflow-gpu 1.7

Download the Yolo model and weight and move to folder “model_data”

How to run:
Step one: Open the file “01-training_corresponding_points.py” and set camera ip. Run the file and the information will be saved in the folder “AutoMS”. To creates the corresponding points follow the steps defined in our paper.

Step two: Open file “02-training_mlp_model.py”, set the path of corresponding points text file and run the python file to train the model.

Step three: the file “03-Reis_et_al._method_versus_your_method.py” you should set the ip cameras and the path of neural network model/weight.

Download

Download the code here.

Contact: Renan Reis (renanreis1@gmail.com)

Reference

If you use the code or parts of it, please cite the following paper.

1.Reis, Renan Oliveira; Dias, Igor Henrique; Schwartz, William Robson (2018): Neural network control for active cameras using master-slave setup. In: International Conference on Advanced Video and Signal-based Surveillance (AVSS), pp. 1-6, 2018. (Type: Inproceeding | Links | BibTeX)