Plant_growth_measurement

Objective: To Detect Line and Caculate the Distance Between Growth Point and Top Flower

Prerequisites and Environment

OS: Ubuntu 18.04

GPU: NVIDIA GeForce RTX 2070

NVIDIA driver version: 470.86

CUDA version: 11.4

How to use this repository

1. Firstly, install anaconda and create your own virtual environment as below.

anaconda version: 4.10.1

conda create -n farmbot python=3.8.5

2. git clone this repository in your workspace.

git clone https://github.com/kistvision/Plant_growth_measurement.git

3. In the virtual environment, farmbot , download the required packages as below.

pip install -r requirements.txt

or pip install the packages described below

pip install numpy==1.22.1
pip install pyrealsense2==2.49.0.3474
pip install torch==1.7.0
pip install pillow==8.3.1
pip install opencv-python==4.5.3.56
pip install Cython==0.29.24
pip install pycocotools==2.0.0
pip install torchvision==0.8.1
pip install matplotlib==3.4.2
pip install PyYAML==5.4.1
pip install pandas==1.3.1
pip install Ipython==7.26.0
pip install tqdm==4.62.0

4. If needed, revise main.py file in order to use your own image samples or lively captured images according to the comments written in main.py.

python main.py --source sample/test --output sample/output --img-size 320 --cfg cfg/yolor_p6_custom.cfg

Or, you can revise main.sh file in order to use above-mentioned parsers more conveniently.

sh main.sh

Brief guideline for using main.sh,

weights: You can download our weights through this Google drive link.

cfg: Our model used yolo_p6_custom.cfg as configuration file.

python main.py \
--weights ./models/best_overall.pt \
--source ./data/testdata \
--output ./outputs \
--conf-thres 0.2 \
--iou-thres 0.5 \
--device 0 \
--agnostic-nms \
--cfg ./yolor/cfg/yolo_p6_custom.cfg\
--names ./data/farmbot.names \
--vertical_focal_len 898.292 \
--horizontal_focal_len 897.507

Result

Input sample image

Result of main.py