Source code of DeepTracking research project
This is an official Torch 7 implementation of the method for the end-to-end object tracking from occluded sensor measurements using neural network presented in the academic paper:
For any questions about the code or the method please contact the author.
Install Torch 7 and the following dependencies (using
luarocks install [package]): * nngraph * image * cunn (optional for training on a GPU)
Download and unzip the training data for the simulated moving balls scenario:
http://mrg.robots.ox.ac.uk:8080/MRGData/deeptracking/DeepTracking_1_1.t7.zipThis is a native Torch 7 file format.
To train the model run:
Training of the neural network using provided data takes about 12 hours on Nvidia Titan X. Every 1000 iterations the training error is logged to log_model.txt, network weights are saved to weights_model and the visualisation of its performance is stored to video_model.
|-gpu [id]||use GPU id|
|-model [file]||neural network model|
|-data [file]||data for training|
|-iter [number]||the number of training iterations|
|-N [number]||the length of training sequences|
|-learningRate [number]||learning rate|
|-initweights [file]||initial weights|
|-grid[minX/maxX/minY/maxY/step] [number]||2D occupancy grid parameters|
|-sensor[start/step]||1D depth sensor parameters|
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.