Commit 03a98701 authored by Jong, Stefan1 de's avatar Jong, Stefan1 de
Browse files

Update README.md

parent 487ffe53
......@@ -172,7 +172,10 @@ When using this code please cite:
## Future research
For future research it can be of interest to use the Tensorflow Object Detection API in combination with a counting line (see image) to create a real time counter. The necessary Tensorflow annotation format (.tfrecord) can be downloaded [here][Tensorflow].
### Optical flow
For future research it can be of interest to implement optical flow within TrackR-CNN. Two data loaders are used during optical flow implementation in TrackR-CNN: `load_optical_flow` & `open_flow_png_file`. A predefined file format is used to calculate minimal flow values in x and y direction. E.g. _000194_x_minimal5.png_ describes the flow in x direction for frame 194 where the minimum flow was 5. _000369_y_minimal-11.png_ is the flow in y direction for frame 369 with a minimal flow of -11. The function `open_flow_png_file` deals with opening the png files and adding back the minimal values as given in the filename. So, when you have some flow files (computed e.g. using Pwcnet), you must first compute the minimum value per file, subtract it from the values for all pixels in that file and then store the result as unsigned 16bit integer .png file. (or you could rewrite `open_flow_png_file` to fit your format). Then `load_optical_flow` can be used for loading the information to establish tracking. Two example optical flow images can be downloaded [here][Opticalflow].
### Tensorflow Object Detection API
For using a simple bounding box counting method the Tensorflow Object Detection API in combination with a counting line (see image) can be applied to create a real time counter. The necessary Tensorflow annotation format (.tfrecord) can be downloaded [here][Tensorflow].
......@@ -194,3 +197,4 @@ For questions related to the code please contact stefan.dejong@wur.nl or create
[CVAT_tracks]: https://cvat.org/documentation/user_guide.html#track-mode-basics
[PP]: https://git.wur.nl/said-lab/rt-obj-tracking/-/tree/master/pre_processing
[CONDA]: https://docs.anaconda.com/anaconda/install/
[Opticalflow]: https://drive.google.com/file/d/1gOPGxalyi7vNL6GSardhYFxZQw6VNoXo/view?usp=sharing
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment