martin-danelljan / eco Goto Github PK

View Code? Open in Web Editor NEW

611.0 33.0 247.0 7.95 MB

Matlab implementation of the ECO tracker.

License: GNU General Public License v3.0

MATLAB 64.16% C++ 23.34% C 12.50%

visual-tracking tracking deep-learning real-time

eco's Introduction

News: Check our pytracking repository, which includes a PyTorch-based implementation of ECO and our most recent trackers ATOM (CVPR 2019) and DiMP (ICCV 2019).

ECO

Matlab implementation of the Efficient Convolution Operator (ECO) tracker.

Publication

Details about the tracker can be found in the CVPR 2017 paper:

Martin Danelljan, Goutam Bhat, Fahad Khan, Michael Felsberg.
ECO: Efficient Convolution Operators for Tracking.
In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017.

Please cite the above publication if you use the code or compare with the ECO tracker in your work. Bibtex entry:

@InProceedings{DanelljanCVPR2017,
Title = {ECO: Efficient Convolution Operators for Tracking},
Author = {Danelljan, Martin and Bhat, Goutam and Shahbaz Khan, Fahad and Felsberg, Michael},
Booktitle = {CVPR},
Year = {2017}
}

Project Webpage

http://www.cvl.isy.liu.se/research/objrec/visualtracking/ecotrack/index.html

Contact

Martin Danelljan

Installation

Using git clone

Clone the GIT repository:

$ git clone https://github.com/martin-danelljan/ECO.git
Clone the submodules.
In the repository directory, run the commands:

$ git submodule init
$ git submodule update
Start Matlab and navigate to the repository.
Run the install script:

|>> install
Run the demo script to test the tracker:

|>> demo_ECO

Note:
This package requires matconvnet [1], if you want to use deep CNN features, and PDollar Toolbox [2], if you want to use HOG features. Both these externals are included as git submodules and should be installed by following step 2. above.

Without using git

You could also downlad and install without using git. This is however not recommented since it will be harder to incorporate updates and you will not get the correct versions of matconvnet and PDollar Toolbox.

Download ZIP file from https://github.com/martin-danelljan/ECO and unpack it somewhere.
Download matconvnet ZIP file from https://github.com/vlfeat/matconvnet and unpack it in the external_libs/matconvnet/ folder of the repository.

Download PDollar Toolbox ZIP file from https://github.com/pdollar/toolbox and unpack it in the external_libs/pdollar_toolbox/ folder of the repository.

Lastly, perform steps 3. and 4. above.

Description and Instructions

Runfiles

The files in the runfiles/ directory are uset to set parameters and run the tracker. You can create your own runfiles by copying an existing one and then play around with different parameters and feature combinations.

These runfiles are included:

OTB_DEEP_settings.m - Settings in the paper used for the deep feature version of ECO on the OTB, UAV123 and TempleColor datasets.
OTB_HC_settings.m - Settings in the paper used for the hand-crafted feature version of ECO (ECO-HC) on the OTB, UAV123 and TempleColor datasets.
VOT2016_DEEP_settings.m - Settings in the paper used for the deep feature version of ECO on the VOT2016 dataset.
VOT2016_HC_settings.m - Settings in the paper used for the hand-crafted feature version of ECO (ECO-HC) on the VOT2016 dataset.
testing_ECO.m - Demo file which contains the same settings as OTB_DEEP_settings.m by default.
testing_ECO_gpu.m - Same as testing_ECO, but uses GPU.
testing_ECO_HC.m - Demo file which contains the same settings as OTB_HC_settings.m by default.
SRDCF_settings.m - Settings that roughly correspond to our previous SRDCF tracker [4] from 2015. The main difference is the optimization method.

Tracking performance on the OTB-2015 dataset is shown bellow for different settings. For comparison, results of our previous trackers C-COT [3], SRDCF [4], DeepSRDCF [5] and DSST [6] are included.

Features

This package includes a quite general framework for feature extraction. You can easily incorporate your own features in the same manner by adding a corresponding "get_featureX.m" function.

Currently, four types of features are included:

Deep CNN features. It uses matconvnet [1], which is included as a git submodule in external_libs/matconvnet/. The imagenet-vgg-m-2048 network available at http://www.vlfeat.org/matconvnet/pretrained/ was used in the ECCV paper. You can use other networks, by placing them in the feature_extraction/networks/ folder.
HOG features. It uses the PDollar Toolbox [2], which is included as a git submodule in external_libs/pdollar_toolbox/.
Lookup table features. These are implemented as a lookup table that directly maps an RGB or grayscale value to a feature vector. Currently, Color Names [8] and Intensity Channels [9] are included.
Colorspace features. Currently grayscale and RGB are implemented.

The tracker supports almost any combination of features. Currently, the only limitation is that you can only use deep features from a single network (but you can use several different layers from the same network).

Each feature has its own parameter settings. You can set the cell size for each non-CNN feature independently. ECO does not assume the same cell size for all feature channels. For the CNN features, you can control the cell size by setting an additional down-sampling factor for each layer.

See the runfile testing.m for examples of how to integrate different features. You can uncomment several features at once in the params.t_features cell array.

fDSST Scale Filter

This reposetery also includes an implementation of the optimized scale filter, which was originally proposed in the fDSST [7]. It is here used in the ECO-HC version of the tracker for speeding-up the scale estimation.

GPU Support

GPU support is activated by setting the parameter "params.use_gpu = true" in the runfile. This requires MatConvNet to be compiled with GPU support. If the install script fails, please visit http://www.vlfeat.org/matconvnet/install/ for instructions.

Integration Into OTB

It should be easy to integrate the tracker into the Online Tracking Benchmark [10]. The runfiles supports the OTB interface, so you just have to copy and rename the runfile you want to use and then add the necessary paths (see setup_paths.m).

Integration Into VOT

To integrate the tracker into the Visual Object Tracking (VOT) challenge toolkit [11], check the VOT_integration folder. Copy the configuration file to your VOT workspace and set the path to the ECO reposetory inside it. The tracker now supports the trax protocol, which is necessary for VOT2017 version of the toolkit.

Raw Results

All raw result files used in our CVPR 2017 paper can be found at the project webpage: http://www.cvl.isy.liu.se/research/objrec/visualtracking/ecotrack/index.html

Why Does the Result Change?

Tracking performance may vary slightly on different machines and whether GPU support is activated. This is due to small numerical effects which can accumulate over time (all trackers are affected by this). Generally, this only affects the performance marginally. More significant changes are rare, but can occur in some videos.

Code References

visionml/pytracking: Python (PyTorch) implementation of ECO and general tracking library containing official implementation of our latest trackers ATOM and DiMP

rockkingjy/OpenTracker: C++ Implementation of ECO and other trackers

StrangerZhang/pyECO: Python implementation of ECO using numpy, mxnet and cupy

References

[1] Webpage: http://www.vlfeat.org/matconvnet/
GitHub: https://github.com/vlfeat/matconvnet

[2] Piotr Dollár.
"Piotr’s Image and Video Matlab Toolbox (PMT)."
Webpage: https://pdollar.github.io/toolbox/
GitHub: https://github.com/pdollar/toolbox

[3] Martin Danelljan, Andreas Robinson, Fahad Khan, Michael Felsberg.
Beyond Correlation Filters: Learning Continuous Convolution Operators for Visual Tracking.
In Proceedings of the European Conference on Computer Vision (ECCV), 2016.
http://www.cvl.isy.liu.se/research/objrec/visualtracking/conttrack/index.html

[4] Martin Danelljan, Gustav Häger, Fahad Khan, Michael Felsberg.
Learning Spatially Regularized Correlation Filters for Visual Tracking.
In Proceedings of the International Conference in Computer Vision (ICCV), 2015.
http://www.cvl.isy.liu.se/research/objrec/visualtracking/regvistrack/index.html

[5] Martin Danelljan, Gustav Häger, Fahad Khan, Michael Felsberg.
Convolutional Features for Correlation Filter Based Visual Tracking.
ICCV workshop on the Visual Object Tracking (VOT) Challenge, 2015.
http://www.cvl.isy.liu.se/research/objrec/visualtracking/regvistrack/index.html

[6] Martin Danelljan, Gustav Häger, Fahad Khan and Michael Felsberg.
Accurate Scale Estimation for Robust Visual Tracking.
In Proceedings of the British Machine Vision Conference (BMVC), 2014.
http://www.cvl.isy.liu.se/research/objrec/visualtracking/scalvistrack/index.html

[7] Martin Danelljan, Gustav Häger, Fahad Khan, Michael Felsberg.
Discriminative Scale Space Tracking.
Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2017.
http://www.cvl.isy.liu.se/research/objrec/visualtracking/scalvistrack/index.html

[8] J. van de Weijer, C. Schmid, J. J. Verbeek, and D. Larlus.
Learning color names for real-world applications.
TIP, 18(7):1512–1524, 2009.

[9] M. Felsberg.
Enhanced distribution field tracking using channel representations.
In ICCV Workshop, 2013.

[10] Y. Wu, J. Lim, and M.-H. Yang.
Online object tracking: A benchmark.
In CVPR, 2013.
https://sites.google.com/site/trackerbenchmark/benchmarks/v10

[11] http://votchallenge.net/

eco's People

Contributors

Stargazers

Watchers

Forkers

bit1002lst andudu damax18 star1322 baiyancheng20 statml benjamesbabala zhengzhugithub xchenjh gzpan xhwxd prozyy ouya-bytes tmomonono zzynwu anatolix wangjuenew caozhengquan sxjscience sltudou4 lifeng9472 silasxue dengyu9080 cybertk opencvfun remgao tjddus9597 arqlalala wsyjwps1983 sunxingxingtf johnkeating24 kixiang alice-ren perception-slam abrams90 forandwhile fxxxysh hxl1990 zhouyzzz hxkwindwizard qunxiang yyuzhongpv machprog hyfine xjsxujingsong ccsuzjj moorejee xiaodongyichuan guangshengshi zh1049 liaoheping zhyj3038 december-boy vbillys dengshuo wjgaas haryslee littlelittlespring kwan-ywan baucheng yueluhongfeng zhanglamei littlecherry11 defypp alakia walkoncross greenteahua gaojie0105 signalimagecv zhangxgu ckldan520 wavesflag mengleijin1 ghgggg biyaa c2h5ohlife arielwanana panxipeng zjuxwg0401 ilikelucifer aa12356jm tigercouple wantajob a2824256 abdelpakey hengshan123 column6942 juliakamelina serycjon xiuxiuzhang1995 alexliyang hong123123 v-italy bigmaye custrep ziliaogithub wangxi-lab alei089 pengboxiangshang mathlee

eco's Issues

How to run the demo_ECO_gpu.m without errors?

Hi, when i run the demo_ECO_gpu.m, errors will occured. but i do not know how to configure related parameters, can anyone help me?

incorrect usage: traxclient Tracker process not alive anymore

what's the matlab version? and other environment

what's the matlab version? and other environment details?

Can ECO be used for multiple object tracking?

Hi,dalao MD :
I`m sorry to disturb, but I just wondering how to init the tracker for multiple object tracking in one video in your code.

How to deploy ECO into cloud

Hello,

thanks for your excellent work.

My question is how to deploy an instance of ECO onto a cloud service like AWS or azure ? Is this even possible with Matlab code ?

Thanks for your help !
Alexander

What's the cnn parameters down_sample for?

Hi, Dear DM. Lately I'm changing VGG in to ResNet for better feature representation, but a little bit confused about how will the downsample parameters influence the outcome. I can't find such details in your paper, could you please give me some intuitive explanations? Much Thanks.

Ps: Using the original VGG, layers[3 14], factorized convolution [16 64], but change downsample into [1 1], I get better outcome(now Precision of OPE 70.9 & original 70.7, Success of OPE now 61.4 & original 61.3).

About the git submodule update

Hello,@martin-danelljan
when I come to the git submodule update,it will have a error:
Please make sure you have the correct access rights
and the repository exists.
fatal: clone of '[email protected]:vlfeat/matconvnet.git' into submodule path 'C:/Users/cpp/ECO/e xternal_libs/matconvnet' failed
Failed to clone 'external_libs/matconvnet' a second time, aborting

then I change the url in gitmudule ,and when I git submodule update again,it still have a error:
fatal: not a git repository: C:/Users/cpp/ECO/external_libs/pdollar_toolbox/../../.git/modules/external_libs/pdollar_toolbox
Failed to clone 'external_libs/pdollar_toolbox'. Retry scheduled
BUG: submodule considered for cloning, doesn't need cloning any more?
Submodule path 'external_libs/matconvnet': checked out '6221423ac64067426bbaa2ffd31557523258d1fe'
fatal: not a git repository: C:/Users/cpp/ECO/external_libs/pdollar_toolbox/../../.git/modules/external_libs/pdollar_toolbox
Unable to fetch in submodule path 'external_libs/pdollar_toolbox'
fatal: not a git repository: C:/Users/cpp/ECO/external_libs/pdollar_toolbox/../../.git/modules/external_libs/pdollar_toolbox
fatal: not a git repository: C:/Users/cpp/ECO/external_libs/pdollar_toolbox/../../.git/modules/external_libs/pdollar_toolbox
fatal: not a git repository: C:/Users/cpp/ECO/external_libs/pdollar_toolbox/../../.git/modules/external_libs/pdollar_toolbox
Fetched in submodule path 'external_libs/pdollar_toolbox', but it did not contain 1a3c9869033548abb0c7a3c2aa6a7902c36f39c2. Direct fetching of that commit failed.

I don't know why,can you help me?

about the “-1” in ‘ECO/utils/get_sequence_info.m’

dear ；
I'm sorry to disturb, I do not understand why to ask for the center position must to -1,and it has much influence to the results.
I am looking forward your answer.

Is there any implementation of ECO in C++ and Opencv?

Is there any implementation of ECO in C++ and Opencv?
There is too many differences between Matlab and C++.
Who has been done it？

Is ECO suitable for feature point tracking?

Hi ! I am curious about whether ECO can be used for feature point tracking. Could you please give some instructions on how to use it for feature point tracking if possible? Thanks!

About the ECO_HC tracker speed problem.

Hi MD,
Thanks for you great work! One thing I need your help, I tested your code on my PC, which is of 24 GB memory, Intel i7-6700 CPU with 3.40GHz x 8, and NVIDIA TITAN X.
When I run ECO_HC file, the speed is only about 20 fps - 30fps, but in the paper said it can run at 60 fps, There are big difference between the two, I want to know why it happened and how should I do to run it at 60 fps.
Sorry for disturbing you and hope you can help me with it.
Thanks!

Error when using mexResize

I have already run compile in external_libs/mexResize
But when I try to use mexReize, there is an ERROR:

Invalid MEX-file '/mydir/data/ECO-master/external_libs/mexResize/mexResize.mexa64':
'/mydir/data/ECO-master/external_libs/mexResize/mexResize.mexa64' is not a valid shared library..

I tried !ldd mexResize in matlab, and the output was:

./mexResize.mexa64: /usr/local/Mathworks/R2017b/sys/os/glnxa64/libstdc++.so.6: version `GLIBCXX_3.4.21' not found (required by ./mexResize.mexa64)
linux-vdso.so.1 => (0x00007ffdba523000)
libopencv_core.so.2.4 => not found
libopencv_imgproc.so.2.4 => not found

I'm not sure what should I do now to run the mexResize, could you help me please?

Can I use live video stream to track?

Hi Martin,

I use testing_ECO_HC.m to test my datasets , but I want combine with detectors to do real-time object detection & tracking . Can you give me some suggestions ?

Thanks in advance

C++ implementation of ECO

Your work is really great, and I am implementing your ECO using C++, try to speed up as much as possible, but still have a question of why you arrange your feature matrix like that in non-GPU version with lots of rearrangements of the order, is that for speed? GPU version matrix is quite clear.

Implementation in python ?

@martin-danelljan the code runs perfect!
Is there any implementation of ECO in python?
If no, how is the workload?
It is hard for me to estimate the workload due to the additional Toolboxes. Thanks!

problem about ECO test on OTB100 !

Hi!
Thanks for you execllent work!I test ECO_HC on OTB100,the pression plot of OPE is normal, but the success plot of OPE is much too high. It's weird. I don't know why, can youn give me some hint?

why not linux?

Does this project support windows only and can it run under Linux?

Compiling errors

Hey when I run the installation script it says:

install

... Build routine for mtimesx
... Checking for PC
building linux version
Building with 'gcc'.
Warning: You are using gcc version '4.8.4'. The version of
gcc is not supported. The version currently supported with
MEX is '6.3.x'. For a list of currently supported compilers
see:
https://www.mathworks.com/support/compilers/current_release. 
> In mtimesx_build (line 111)
  In install (line 10)

any help would be great!

problem about testing ECO on OTB-2015

Hi, Martin,

Your work is really, really great, and I want to ask some questions.
In your paper, the success plot of OPE of ECO is 0.700, ECO-HC is 0.650, C-COT is 0.690, Staple is 0.584. But when I test these modles on OTB-2015, I get the result ECO is 0.688, ECO-HC is 0.635, C-COT is 0.671, Staple is 0.582. You can see that my result is a bit lower than yours. Could you give me some suggestions to increase those values?

Thank you very much.

some questions about the DSST

the code:
im_patch_scale = getScaleSubwindow(im, pos, base_target_sz, scale_factor*scale_factors, scale_window, scale_model_sz, p.hog_scale_cell_size); xsf = fft(im_patch_scale,[],2);

why the 33 scales multiply the scale_window(shape is 1*33) respectively? It means the farther away 'base_target_sz' the maller probability? Especially, scale_window(1) = scale_window(end) = 0, why don't delete these two scales?

fft(X, [], 2), the arg '2' it means fft by row, but we konw each column of 'im_patch_scale' is a scale. so, why don't set the arg '1' to fft by column for a scale image? PLS.

About result on OTB with ECO-HC

Hello,

I meet a problem when I run the ECO-HC. The success plots and precision that I got from OTB-100 and OTB-50 are 3% lower than the results in the paper.I don’t know what’s wrong.

Can you help me!

QiujieDong

test on vot2017

hi Martin,

I'm trying to run_experiment on vot2017 using trax, but it replied 'don't get response'. Maybe some slightly detailed mismatches for the trax protocol?

Thanks for your reply in advance.

best,
Lichao

ECO tracking source code issue

About results on VOT2016 with ECO-HC

Hello ,DM

Thank you for your excellent working. I run the experiments on VOT2016 recently but got strange results. I use the VOT_integration/configuration_file/tracker_ECO.m directly and got 0.3579, the baseline EAO， which is the same as that in #1. And when I use VOT2016_HC_settings, I got 0.28 only, which is much worse than 0.32 in your paper. Can different machine cause so different results? Is there any other reason lead to that?

The sub-module that generates the groundtruth_rect.txt

Hello!

I'm using your code on a school project. Recently I ran the demo_ECO and found that the target was chosen in advance. So I wonder if there exists a sub-module that generates the file
"groundtruth_rect.txt" so that I can choose an object in the first frame and track it.

Lot of thanks in advance!

when i integrate the eco with vot toolkits(5.0.2) there some error

use the file you provide
and if i set the repetition of the realtime is 1,there is no problem
when i debug this, i find the T1 and T2 is empty
maybe i could spend alot time on the vot toolkit to understand the problem
could you help me about this ?

the robusteness when the the person object turn round?

Hi, sorry to disturb you. how about the robusteness when the the person object turn round? @martin-danelljan @goutamgmb @cybertk

About some problems of sh

sh = mtimesx(samplesf{k1}, permute(hf{k1}, [3 4 1 2]), 'speed');
sh(:,1,1+pad_sz{k}(1):end-pad_sz{k}(1), 1+pad_sz{k}(2):end) = ...
sh(:,1,1+pad_sz{k}(1):end-pad_sz{k}(1), 1+pad_sz{k}(2):end) + mtimesx(samplesf{k}, permute(hf{k}, [3 4 1 2]), 'speed');

hello:
Your work is great.Recently,I plan to implement your ECO using C++,but I met some problems about the variable ''sh'',I have a question of what it denotes in the lhs_operation_joint.m,and I wonder why you occur to implement the two steps.

About Multi-tracking Problem

HI @martin-danelljan ,

Thanks for your impressive work! I just successfully run the demo_ECO_HC in matlab 2014b and it seems perfect. However, I notice that it only tracks one person, so I am curious about the performance in multi-tracking task.

Also, I am interested in using this tracker in other sequences. I am not sure but it seems that your demo is just read the frame and bbox-info in 'sequence' file? So how can I try it by using my own sequences? Quite confuse when I find out the 'tracker.m' but cannot figure out ways to import sequences... I am a newbie in using Matlab ...

Thx!

About sample weights and scale estimation

I would like to know why CCOT and ECO do not use adaptive decontamination strategy as SRDCF_decon do and scale estimation as fDSST do? Isn't SRDCF_decon an effective method to solve the problem of model update? Though scale estimation strategy from fDSST is included in ECO, it is not used in OTB and VOT test. I am confusing.

Hi, MD

installation of submodules

Hi, I ran into the issues below while trying to clone the external libraries for ECO. It appears to be a failure to clone the pdollar and matconvnet libraries. Do you have any ideas to help me figure out what I should do?

C:\Users\nkiruka\Desktop\tracking>cd ECO

C:\Users\nkiruka\Desktop\tracking\ECO>git submodule init
Submodule 'external_libs/matconvnet' ([email protected]:vlfeat/matconvnet.git) registered for path 'external_libs/matconvnet'
Submodule 'external_libs/pdollar_toolbox' ([email protected]:pdollar/toolbox.git) registered for path 'external_libs/pdollar_toolbox'

C:\Users\nkiruka\Desktop\tracking\ECO>git submodule update
Cloning into 'C:/Users/nkiruka/Desktop/tracking/ECO/external_libs/matconvnet'...
Host key verification failed.
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.
fatal: clone of '[email protected]:vlfeat/matconvnet.git' into submodule path 'C:/Users/nkiruka/Desktop/tracking/ECO/external_libs/matconvnet' failed
Failed to clone 'external_libs/matconvnet'. Retry scheduled
Cloning into 'C:/Users/nkiruka/Desktop/tracking/ECO/external_libs/pdollar_toolbox'...
Host key verification failed.
fatal: Could not read from remote repository.

Please make sure you have the correct access rights
and the repository exists.
fatal: clone of '[email protected]:pdollar/toolbox.git' into submodule path 'C:/Users/nkiruka/Desktop/tracking/ECO/external_libs/pdollar_toolbox' failed
Failed to clone 'external_libs/pdollar_toolbox'. Retry scheduled
Cloning into 'C:/Users/nkiruka/Desktop/tracking/ECO/external_libs/matconvnet'...
Host key verification failed.
fatal: Could not read from remote repository.

Error when i run vl_testnn('gpu',true)?

         Bus error detected at Fri Nov 16 20:06:56 2018

Configuration:
Crash Decoding : Disabled - No sandbox or build area path
Crash Mode : continue (default)
Current Graphics Driver: Unknown hardware
Current Visual : 0x47 (class 4, depth 24)
Default Encoding : UTF-8
Deployed : false
GNU C Library : 2.23 stable
Host Name : fyj-B85-HD3
MATLAB Architecture : glnxa64
MATLAB Entitlement ID: 6257193
MATLAB Root : /usr/local/MATLAB/R2016b
MATLAB Version : 9.1.0.441655 (R2016b)
OpenGL : hardware
Operating System : Linux 4.15.0-39-generic #42~16.04.1-Ubuntu SMP Wed Oct 24 17:09:54 UTC 2018 x86_64
Processor ID : x86 Family 6 Model 60 Stepping 3, GenuineIntel
Virtual Machine : Java 1.7.0_60-b19 with Oracle Corporation Java HotSpot(TM) 64-Bit Server VM mixed mode
Window System : The X.Org Foundation (11906000), display :0

Fault Count: 1

Abnormal termination:
Bus error

About "report_tracking_results.m"

Does this function work on results and sequence obtained using any method proposed in the repository?
I think it does not.

when I run demo_ECO.m ，there is a question，can you help me？

The error was:
MEX-file '/home/zl/software/ECO/external_libs/mexResize/mexResize.mexa64' is useless:
lack of '/usr/local/lib/libopencv_core.so.2.4->/home/zl/software/ECO/external_libs/mexResize/mexResize.mexa64' symbol
'_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE12find_last_ofEPKcmm'
lack of '/usr/local/lib/libopencv_core.so.2.4->/home/zl/software/ECO/external_libs/mexResize/mexResize.mexa64' symbol
'_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE17find_first_not_ofEPKcmm'
lack of '/usr/local/lib/libopencv_core.so.2.4->/home/zl/software/ECO/external_libs/mexResize/mexResize.mexa64' symbol
'_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE4findEcm'
lack of '/usr/local/lib/libopencv_core.so.2.4->/home/zl/software/ECO/external_libs/mexResize/mexResize.mexa64' symbol
'_ZNKSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEE4findEPKcmm'

I have already installed opencv2.4.13 in my ubuntu before，Is opencv path conflict?

Install.m Error

https://i.imgur.com/6sZzPM9.png

i am getting this error when i am trying to run install.m

'cl.exe' is not recognized as an internal or external command,
operable program or batch file.
Error using vl_compilenn>check_clpath (line 599)
Unable to find cl.exe

Error in vl_compilenn (line 417)
cl_path = fileparts(check_clpath()); % check whether cl.exe in path

Error in install (line 33)
vl_compilenn;

When i run install.m ,Wrong:“Error using mex incorrectly”

Hello!
Now,I'm using your code to learn tracking. following your methods, when I run the code of install.m, prompting an error ,this ：
ECO-master\external_libs\matconvnet\matlab\src\bits\impl/dispatcher.hpp:50:16: error: request for member 'tensor' in 'h',which is of non-class type 'int'
if ((int)h.tensor && !(int)(h.tensor->isNull())) {

Error when I use combination of CNN and CN

Hi! I wanted to try a different combination of features from your paper. So I copied .\runfiles\VOT2016_DEEP_settings.m and changed something.

After I commented struct('getFeature',@get_fhog,'fparams',hog_params),... without changing any other things and run run_test.m, it reported 'Tracker execution interrupted: Unable to establish connection.' Do you know the reason?

Error when I run demo_ECO_gpu.m？

I have succeed to run the demo_ECO.m and demo_ECO_HC.m file. However, I have encountered these errors.
Error using vl_nnconv
An input is not a numeric array (or GPU support not compiled).

Error in vl_simplenn (line 300)
res(i+1).x = vl_nnconv(res(i).x, l.weights{1}, l.weights{2}, ...

Error in get_cnn_layers (line 23)
cnn_feat = vl_simplenn(fparams.net, im,[],[],'CuDNN',true, 'Mode', 'test');

Error in extract_features (line 59)
feature_map(ind:ind+num_blocks-1) = feat.getFeature(img_samples{img_sample_ind},
feat.fparams, gparams);

Error in tracker (line 347)
xl = extract_features(im, sample_pos, currentScaleFactor, features, global_fparams,
feature_extract_info);

Error in testing_ECO_gpu (line 129)
results = tracker(params);

Error in demo_ECO_gpu (line 13)
results = testing_ECO_gpu(seq);

Could you help to fix these problems?

Add equation of EAO

Hi Martin,

Where should i add the equation of expected average overlap (EAO) into the matlab code?

Thanks
Best regards,
Lylee

Install error

Hi there, I followed every step but still got this error when I type "install' in matlab.
I'm sure I did these two lines
$ git submodule init
$ git submodule update

Does anyone know how to deal with it? It seems to be something wrong with mtimesx
Thanks a lot!

Windos10, Matlab R2016a

... Build routine for mtimesx
... Checking for PC
... Finding path of mtimesx C source code files
... Found file mtimesx.c in C:\Users\32621\Desktop\ECO\external_libs\mtimesx\mtimesx.c
... Found file mtimesx_RealTimesReal.c in C:\Users\32621\Desktop\ECO\external_libs\mtimesx\mtimesx_RealTimesReal.c
错误使用 mtimesx_build (line 166)
A C/C++ compiler has not been selected with mex -setup

出错 install (line 10)
mtimesx_build;

About results on VOT2016

Hello ,DM

I have run the experienment on VOT2016 for several hours. And the EAO is 0.3581 on baseline , lower than 0.375 in the paper . I doubt whether the param causes the difference ?

About ECO tracking results on VOT2016

Hi! I just tested the ECO tracker on the VOT2016 dataset but the result curve below is different from the ECO paper. Is there something wrong with my testing results?

the project didn't work for vot2016 dataset?

hi, I download the vot2016 datasets, place some video in the directory sequences, run the demo, the program doesn't work. and I find the program has the setting: 'seq.format = otb,' I change it to 'vot', it does't work. how can I change the program? thx @cybertk @goutamgmb @martin-danelljan

Testing VOT2016 Evaluation yields worse results

Hello Martin, first I wanted to thank you for an amazing job you have made alongside your peers.

As the title said, I'm getting different results compared to what you showed in the ECO paper.
My first tests were done using the ECO version available at the VOT Challenge website, and obtained the following results:

EAO Values 1st try

After seeing those results, I decided to check the code in this GitHub, and found there was a minor difference with the VOT version.
In the code downloaded from VOT the "learning_rate" param was set to "0.011", and in this repo it is set to "0.012". The other difference was the presence of a "neglect_higher_frequency" parameter that is not present in the code from this repo.
After applying those changes, I run the evaluation and analysis again and obtained the following results:

EAO Values 2nd try

I still couldn't reproduce the same results. Why do you think this can be happening and end up with a 0.04 EAO difference?

For the test I used the followings

Windows 10 platform
VOT Toolkit v6.0.3
Matlab R2017a
MatConvNet 1.0-beta24

Again, thanks for the hard work.
Best regards, Lucas.

demo_ECO no work

Error using fliplr (line 18)
X must be a 2-D matrix.
Error in cfft2 (line 19)
xf(end,:,:,:) = conj(fliplr(xf(1,:,:,:)));
Error in tracker (line 297)
xlf = cellfun(@cfft2, xlw, 'uniformoutput', false);
Error in testing_ECO (line 133)
results = tracker(params);
Error in demo_ECO (line 13)
results = testing_ECO(seq);

could you released the code of this picture ?

thx advance.

Some questions about the "Scale filter"

Hi, when I compare the 'fdsst' code with 'dsst' code， I find that

The value of ys is：
[1 0.642167199144790 0.170056201828270 0.0185709050136830 0.000836315026159023 1.55311269123150e-05 1.18941188747878e-07 3.75627879293239e-10 4.89192285304428e-13 4.89192285304428e-13 3.75627879293239e-10 1.18941188747878e-07 1.55311269123150e-05 0.000836315026159023 0.0185709050136830 0.170056201828270 0.642167199144790]
The largest value is the first value.

However in DSST code, the largest value is in the middle. As to the transformation filter, the largest value is also different in the both codes. Any suggestion would be appreciated.

```

scale_exp = (-floor((nScales-1)/2):ceil((nScales-1)/2)) * nScalesInterp/nScales;
scale_exp_shift = circshift(scale_exp, [0 -floor((nScales-1)/2)]);

interp_scale_exp = -floor((nScalesInterp-1)/2):ceil((nScalesInterp-1)/2);
interp_scale_exp_shift = circshift(interp_scale_exp, [0 -floor((nScalesInterp-1)/2)]);

scaleSizeFactors = scale_step .^ scale_exp;
interpScaleFactors = scale_step .^ interp_scale_exp_shift;

ys = exp(-0.5 * (scale_exp_shift.^2) /scale_sigma^2);

How to train ECO model?

Hello! How do you train the ECO model? Can you tell me some information?

CPU only

Hello, I want to know whether can I use this code without GPU?