sushlokshah/optical_flow

Optical Flow

setup:

1. Download Kitti Optical Flow Dataset
2. git clone https://github.com/sushlokshah/optical_flow.git
3. cd optical_flow/
4. mkdir dataset
5. extract the dataset in this folder

   directory structure
       |-- dataset
           |-- data_scene_flow
               |-- testing
               |-- training
   

6. mkdir results
7. go the root directory and download dependencies

conda env create -f environment.yml

8. create executables of cpp code

cd evaluation/cpp
mkdir build
cd build
cmake ..
make

9. 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