- Download Kitti Optical Flow Dataset
- git clone https://github.com/sushlokshah/optical_flow.git
- cd optical_flow/
- mkdir dataset
- extract the dataset in this folder
directory structure |-- dataset |-- data_scene_flow |-- testing |-- training
- mkdir results
- go the root directory and download dependencies
conda env create -f environment.yml
- create executables of cpp code
cd evaluation/cpp
mkdir build
cd build
cmake ..
make
- For estimating the results: execute the python file with the appropriate method. for example.
python flow.py -m Farneback_flow
-
Hackmd link: https://hackmd.io/@Sushlok/S1FMH1Wv5/edit
-
kitti result format
- num_errors_bg/num_pixels_bg
- num_errors_bg_result/num_pixels_bg_result
- num_errors_fg/num_pixels_fg
- num_errors_fg_result/num_pixels_fg_result
- num_errors_all/num_pixels_all
- num_errors_all_result/num_pixels_all_result
- Results: Sheet_link
- reviewed papers: drive link