This tracker is KCF based one with CNN features, and this version is made for both OTB and VOT evaluation.
*** Most Inportant, Do use the tensorflow version 1.1 and Cudnn version 5.1, because the higher version will have conflicts in hdf5, which haven't been solved by now.***
- Download inception_v3.ckpt from Kaggle, and link to the current holder
- test the deep kcf by running run_deepkcy.py
* update tensorflwo into r1.1 `sudo pip install --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.1.0-cp27-none-linux_x86_64.whl`
* use cudnn 5.1, download cudnn 5.1 `https://developer.nvidia.com/rdp/cudnn-download`, and install under the cudnn install direction `http://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html`.
git clone git@github.com:ugv-tracking/OTB_Tracker.git
cd OTB_Tracker/trackers
git clone git@github.com:ugv-tracking/DeepKCF.git
open Matlab and change the dir to OTB_Tracker
run main_running.m
check util/configSeqs.m
check util/configTrackers.m
run perfPlot.m
git clone git@github.com:ugv-tracking/VOT-Tracker.git
cd VOT-Tracker/workspace_DKCF
git clone git@github.com:ugv-tracking/DeepKCF.git
cd VOT-Tracker/workspace_DKCF
run_experiments.m % This script can be used to execute the experiments for a single tracker
run_test.m % This script can be used to test the integration of a tracker to the framework.
run_analysis.m % This script can be used to perform a comparative analyis of the experiments in the same manner as for the VOT challenge.
Check the below files:
- tracker.py
- DeepKCF_OTB.py
- DeepKCF_VOT.py