In agriculture, automating the accurate tracking of fruits, vegetables, and fiber is a very tough problem. The issue becomes extremely challenging in dynamic field environments. Yet, this information is critical for making day-to-day agricultural decisions, assisting breeding programs, and much more.
This repository provides source code for our 2023 IEEE Transactions on Automation Science and Engineering article titled "NTrack: A Multiple-Object Tracker and Dataset for Infield Cotton Boll Counting." NTrack, is a multiple object tracking framework based on the linear relationship between the locations of neighboring tracks. It computes dense optical flow and utilizes particle filtering to guide each tracker. Correspondences between detections and tracks are found through data association via direct observations and indirect cues, which are then combined to obtain an updated observation. NTrack is independent of the underlying detection method, thus allowing for the interchangeable use of any off-the-shelf object detector. NTrack was created for the task of tracking and counting infield cotton bolls. To develop and test NTrack, we created TexCot22, an infield cotton boll video dataset. Each tracking sequence was collected from unique rows of an outdoor cotton crop research plot located in the High Plains region of Texas.
More information on the project can be found on the NTrack website.
If you find this project useful, then please consider citing both our paper and dataset.
@article{muzaddid2023ntrack,
title={NTrack: A Multiple-Object Tracker and Dataset for Infield Cotton Boll Counting},
author={Muzaddid, Md Ahmed Al and Beksi, William J},
journal={IEEE Transactions on Automation Science and Engineering},
volume={},
number={},
pages={1--13},
doi={10.1109/TASE.2023.3342791},
year={2023}
}
@data{T8/5M9NCI_2024,
title={{TexCot22}},
author={Muzaddid, Md Ahmed Al and Beksi, William J},
publisher={Texas Data Repository},
version={V2},
url={https://doi.org/10.18738/T8/5M9NCI},
doi={10.18738/T8/5M9NCI},
year={2024}
}
First, begin by cloning the project:
$ git clone https://github.com/robotic-vision-lab/NTrack-A-Multiple-Object-Tracker.git
$ cd NTrack-A-Multiple-Object-Tracker
Next, create an environment and install the dependencies:
$ conda create --name ntrack python=3.8.
$ conda activate ntrack
$ pip install -r requirements.txt
Download the TexCot22 dataset files. TexCot22 consists of five ZIP files for tracking (TexCot22-[1-5]). Each ZIP file contains training and testing sequences, ground-truth bounding boxes, and the detections. The dataset directories have the following structure:
TexCot22-X/
- train
- vid
- img1
- 0001.jpg
- 0002.jpg
- ...
- gt
- gt.txt
- ...
- test
- vid
- img1
- 0001.jpg
- 0002.jpg
- ...
- det
- det.txt
- gt
- gt.txt
- ...
In addition, there are four ZIP files (TexCot22_Detection-1_[1-2], TexCot22_Detection-2_[1-2]), which can be used to train a detection model. To train an off-the-shelf object detector, unzip TexCot22_Detection-X_1.zip and TexCot22_Detection-X_2.zip into the same directory and then merge the two img directories into one. The file frameid_to_imgfile.npy contains the mapping from the frameid (1st column in detection.csv) to the image name in the img folder.
To run on a test data split, invoke the following command:
$ python ntrack.py --data_base_dir {your/data/dir/TexCot22-X} --data_split test --use_pf True
