-# KCF tracker parallel and PREM implementations
-This project aims to test multiple implementation of calculating Fast Fourier Transform(FFT) and Inverse Fourier Transform(IFFT) in KFC tracker and to see how the performance changes depending on the implementation.
+# KCF tracker – parallel and PREM implementations
-C++ implementation of KCF tracker was written by Tomas Vojir [here](https://github.com/vojirt/kcf/blob/master/README.md) and is reimplementation of algorithm presented in "High-Speed Tracking with Kernelized Correlation Filters" paper[1].
+The goal of this project is modify KCF tracker for use in the
+[HERCULES][1] project, where it will run on NVIDIA TX2 board. To
+achieve the needed performance we try various ways of parallelization
+of the algorithm including execution on the GPU. The aim is also to
+modify the code according to the PRedictable Execution Model (PREM).
-### Prerequisites
-The code depends on OpenCV 3.0+ library and cmake is used for building it. Depending on the implementation selected you have to have installed [FFTW](http://www.fftw.org/), [CUDA](https://developer.nvidia.com/cuda-downloads) or [OpenMP](http://www.openmp.org/).
+Stable version of the tracker is available from a [CTU server][2],
+development happens at GitHub [here][wsh] and [here][3].
-SSE instructions were used in the original code and these are only supported on x86 architecture. Thanks to the [SSE2NEON](https://github.com/jratcliff63367/sse2neon) code, we now support both ARM and x86 architecture.
+[1]: http://hercules2020.eu/
+[2]: http://rtime.felk.cvut.cz/gitweb/hercules2020/kcf.git
+[wsh]: https://github.com/wentasah/kcf
+[3]: https://github.com/Shanigen/kcf
-## Getting Started
-Open terminal in the directory with the code:
-Using cmake gui:
-________________
+## Prerequisites
+
+The code depends on OpenCV 2.4 library
+and [CMake][13] (optionally with [Ninja][8]) is used for building.
+Depending on the version to be compiled you need to have development
+packages for [FFTW][4], [CUDA][5] or [OpenMP][6] installed.
+
+On TX2, the following command should install what's needed:
+``` shellsession
+$ apt install cmake ninja-build libopencv-dev libfftw3-dev
+```
+
+[4]: http://www.fftw.org/
+[5]: https://developer.nvidia.com/cuda-downloads
+[6]: http://www.openmp.org/
+[13]: https://cmake.org/
+
+## Compilation
+
+There are multiple ways how to compile the code.
+
+### Compile all supported versions
+
+``` shellsession
+$ git submodule update --init
+$ make -k
+```
+
+This will create several `build-*` directories and compile different
+versions in them. If prerequisites of some builds are missing, the
+`-k` option ensures that the errors are ignored. This uses [Ninja][8]
+build system, which is useful when building naively on TX2, because
+builds with `ninja` are faster (better parallelized) than with `make`.
+
+To build only a specific version run `make <version>`. For example,
+CUDA-based version can be compiled with:
+
+``` shellsession
+$ make cufft
```
+
+[8]: https://ninja-build.org/
+
+### Using cmake gui
+
+```shellsession
+$ git submodule update --init
$ mkdir build
-$ cmake-gui //Select the directory path and build directory. After which you can choose desired build option.
- //Each option has comment explaining briefly what it does.
-$ make
+$ cmake-gui .
```
-Without cmake gui:
-___________________
+
+- Use the just created build directory as "Where to build the
+ binaries".
+- Press "Configure".
+- Choose desired build options. Each option has a comment briefly
+ explaining what it does.
+- Press "Generate" and close the window.
+
+```shellsession
+$ make -C build
```
+### Command line
+
+```shellsession
+$ git submodule update --init
$ mkdir build
$ cd build
+$ cmake [options] .. # see the tables below
+$ make
```
-The following table shows multiple options how to run cmake to get different version of the tracker.
+The `cmake` options below allow to select, which version to build.
-| Option| Description |
-| --- | --- |
-| `cmake ..` | Single thread version of original project with OpenCV FFT.|
-| `cmake -DASYNC=ON ..` | Multi thread version of original project with OpenCV FFT together with C++ async directive.|
-| `cmake -DOPENMP=ON ..` | Use OpenMP with OpenCV implementation.|
-| `cmake -DFFT=OpenCV_cuFFT ..`**WIP** | Nvidia CUFFT implemented in OpenCV will be used. Together with Hostmem from OpenCV.|
-| `cmake -DFFT=fftw ..`**WIP** | Use FFTW implementation of FFT.|
-| `cmake -DFFT=fftw -DOPENMP=ON ..`**WIP** | Use OpenMP library with FFTW.|
-| `cmake -DFFT=fftw -DASYNC=ON ..`**WIP** | Use C++ async directive with FFTW.|
+The following table shows how to configure different FFT
+implementations.
-To all of these you can also add these additional options:
+|Option| Description |
+| --- | --- |
+| `-DFFT=OpenCV` | Use OpenCV to calculate FFT.|
+| `-DFFT=fftw` | Use fftw and its `plan_many` and "New-array execute" functions. If `std::async`, OpenMP or cuFFTW is not used the plans will use 2 threads by default.|
+| `-DFFT=cuFFTW` | Use cuFFTW interface to cuFFT library.|
+| `-DFFT=cuFFT` | Use cuFFT. This version also uses pure CUDA implementation of `ComplexMat` class and Gaussian correlation.|
-There is also option for profiling. Currently only profiles *get_features* function for which it calculates average number of CPU cycles. Works with default single threaded version using OpenCV. The code was taken from [libpfm4](https://sourceforge.net/p/perfmon2/libpfm4/ci/master/tree/) perf_examples self_basic.c:
+With all of these FFT version additional options can be added:
-| Option| Description |
+|Option| Description |
| --- | --- |
-| `-DPROFILING=ON` | Enable profiling using perf_event together with libpfm4.|
+| `-DASYNC=ON` | Use C++ `std::async` to run computations for different scales in parallel. This doesn't work with `BIG_BATCH` mode.|
+| `-DBIG_BATCH=ON` | Concatenate matrices of different scales to one big matrix and perform all computations on this matrix. This mode doesn't work with `OpenCV` FFT.|
+| `-DOPENMP=ON` | Parallelize certain operation with OpenMP. This can only be used with `OpenCV` or `fftw` FFT implementations. By default it runs computations for differenct scales in parallel. With `-DBIG_BATCH=ON` it parallelizes the feature extraction and the search for maximal response for differenct scales. With `fftw`, Ffftw's plans will execute in parallel.|
+| `-DCUDA_DEBUG=ON` | Adds calls cudaDeviceSynchronize after every CUDA function and kernel call.|
+| `-DOpenCV_DIR=/opt/opencv-3.3/share/OpenCV` | Compile against a custom OpenCV version. |
-Finally call make:
-```
-$ make
-```
-This code compiles into binary **kcf_vot**
+### Compilation for non-TX2 CUDA
+
+The CuFFT version is set up to run on NVIDIA Jetson TX2. If you want
+to run it on different architecture, change the `--gpu-architecture
+sm_62` NVCC flag in **/src/CMakeLists.txt** to your architecture of
+NVIDIA GPU. To find what SM variation you architecture has look
+[here][9].
+
+[9]: http://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/
+
+## Running
+
+No matter which method is used to compile the code, the results will
+be `kcf_vot` binary.
+
+It operates on an image sequence created according to [VOT 2014
+methodology][10]. You can find some image sequences in [vot2016
+datatset][11].
+
+The binary can be run as follows:
+
+1. `./kcf_vot [options]`
+
+ The program looks for `groundtruth.txt` or `region.txt` and
+ `images.txt` files in current directory.
-./kcf_vot [-v[delay\_ms]]
-- using [VOT 2014 methodology](http://www.votchallenge.net/)
-- to get dataset used in VOT go [here](http://www.votchallenge.net/vot2016/dataset.html)
- - INPUT : expecting two files, images.txt (list of sequence images with absolute path) and
- region.txt with initial bounding box in the first frame in format "top_left_x, top_left_y, width, height" or
- four corner points listed clockwise starting from bottom left corner.
- - OUTPUT : output.txt containing the bounding boxes in the format "top_left_x, top_left_y, width, height"
+ - `images.txt` contains a list of images to process, each on a
+ separate line.
+ - `groundtruth.txt` contains the correct location of the tracked
+ object in each image as four corner points listed clockwise
+ starting from bottom left corner. Only the first line from this
+ file is used.
+ - `region.txt` is an alternative way of specifying the location of
+ the object to track via its bounding box (top_left_x, top_left_y,
+ width, height) in the first frame.
-
+2. `./kcf_vot [options] <directory>`
-## Author
-* **Karafiát Vít**
+ Looks for `groundtruth.txt` or `region.txt` and `images.txt` files
+ in the given `directory`.
+
+3. `./kcf_vot [options] <path/to/region.txt or groundtruth.txt> <path/to/images.txt> [path/to/output.txt]`
+
+By default the program generates file `output.txt` containing the
+bounding boxes of the tracked object in the format "top_left_x,
+top_left_y, width, height".
+
+[10]: http://www.votchallenge.net/
+[11]: http://www.votchallenge.net/vot2016/dataset.html
+
+### Options
+
+| Options | Description |
+| ------- | ----------- |
+| --visualize, -v[delay_ms] | Visualize the output, optionally with specified delay. If the delay is 0 the program will wait for a key press. |
+| --output, -o <output.txt> | Specify name of output file. |
+| --debug, -d | Generate debug output. |
+| --fit, -f[W[xH]] | Specifies the dimension to which the extracted patch should be scaled. It should be divisible by 4. No dimension or zero rounds the dimensions to the nearest smaller power of 2, a single dimension `W` will result in patch size of `W`×`W`. |
+
+
+## Authors
+* Vít Karafiát, Michal Sojka
+
+Original C++ implementation of KCF tracker was written by Tomas Vojir
+[here][12] and is reimplementation of algorithm presented in
+"High-Speed Tracking with Kernelized Correlation Filters" paper \[1].
+
+[12]: https://github.com/vojirt/kcf/blob/master/README.md
## References
-[1] João F. Henriques, Rui Caseiro, Pedro Martins, Jorge Batista, “High-Speed Tracking with Kernelized Correlation Filters“,
-IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
-_____________________________________
-Copyright (c) 2014, Tomáš Vojíř
+\[1] João F. Henriques, Rui Caseiro, Pedro Martins, Jorge Batista,
+“High-Speed Tracking with Kernelized Correlation Filters“, IEEE
+Transactions on Pattern Analysis and Machine Intelligence, 2015
+
+## License
+
+Copyright (c) 2014, Tomáš Vojíř\
+Copyright (c) 2018, Vít Karafiát\
+Copyright (c) 2018, Michal Sojka
Permission to use, copy, modify, and distribute this software for research
purposes is hereby granted, provided that the above copyright notice and
projects / hercules2020 / kcf.git / blobdiff