A

Project Presentation
on
Weed And Crop Detection Using ML

Presented By
Vivek (2000971520063)
Rishabh Sonkar(2000971520045)
Vini Srivastava(2000971520062)
Guided By
Mr. Mukesh Kumar Singh

Department of Computer Science & Engineering

Galgotias College of Engineering & Technology , Greater Noida
Dr.A.P.J.Abdul Kalam Technical University, Lucknow, Uttar Pradesh
INTRODUCTION
• Weed detection System is a system that can effectively
improve the efficiency in controlling the weeds without
using a lot of workforce in finding the weeds manually and
spray the herbicides.

• This model differentiate between crop and weed using

YOLOv3 on agriculture image data.
 What is WEED?

• Weeds are undesirable plants that grow in agricultural

crops, competing for elements such as sunlight and
water, causing losses to crop yields

• It makes difficult to operate harvesting machines.

• Several process has been carried out of identification

and classification of weeds
 LITERATURE SURVEY
Author Year Targeted Crop Dataset Model/Tools Accuracy Research Gap
Problem
Definition

Crop weed
Images were collected from descrimination in
Alam [1] 2021 Crop/weed detection and Unspecified private farm RF 95% precision agriculture
classification

Brinkhoff [2] 2020 Land cover mapping 9 perennial crops Images were taken from the SVM 84.80% Weed map
Riverina region in

Robust Crop and Weed

Detection in Precision
Aaron [3] 2019 Weed detection Maize Images were collected from 98% Agriculture
private farm NVDI digit,
YOLOv3 Detector

Zhang [4] 2019 Weeds species recognition 8 Weed plants SVM 92.35% Deep Weeds
1600 Weed images were taken
from South China crop field
Author Year Targeted Crop Dataset Model/Tools Accuracy Research Gap
Problem
Definition

Avocado field
Y-H Tu [5] 2019 Measuring Canopy structure Avocado tree RF 96% Data Annotation and
Bundaberg, Labeling
Australia

Adel [6] 2018 Sugar beet Images were taken by SVM 95% Real-World Testing
Weed detection using shape
feature Shiraz
University, Iran

2018 Weeds species detection Soybean SVM 95.07% Interactions and

Abouzahir [7] Images were collected from Dynamics

Sâa José farm,

Brazil
 RESEARCH GAP
1. Generalization Across Crops: Most research primarily addresses the detection of a specific crop or weed type. There
is a need for research that can generalize detection models to accommodate various crop and weed species,
providing flexibility to adapt to diverse agricultural settings.
2. Data Annotation and Labeling: Creating annotated datasets for training machine learning models in crop and weed
detection is a time-consuming and costly process. Research is required to develop efficient methods for data
annotation and labeling, including semi-supervised and weakly supervised approaches.
3. Real-World Testing: While many algorithms perform well in controlled environments, their performance often
degrades when applied to real-world agricultural settings with complex vegetation, varying weather conditions, and
different soil types. Research should focus on testing and validating detection models in diverse and challenging
field conditions.
4. Interactions and Dynamics: Crop-weed interactions are dynamic and change over time. Research gaps exist in the
development of detection models that can capture the evolving nature of crop-weed interactions throughout the
growing season. Understanding how to adapt detection algorithms to changing crop and weed dynamics is crucial
for precision agriculture.
5. Sparse Point Clouds: Existing research often focuses on high-density point clouds, while sparse point clouds, which
are more common in practical applications due to cost constraints, remain underexplored. Developing robust
algorithms that can effectively detect crops and weeds in sparse point clouds is a significant research gap.
RESEARCH OBJECTIVE
• Learning more about weed control and detection can make it easier for
farmers to decide whether these weeds should be welcomed or if they
must be removed.

• Weed detection and prevention can already begin before planting Weed
seeds could be recognized and removed, minimizing the growth of
unwanted plants.

• Develop and refine image processing and machine learning algorithms

for accurate weed detection in various agricultural settings.

• To minimize the use of fertilizers and decrease the acidity of the soil.
PROBLEM STATEMENT
Weed plant detection is a new research problem in agricultural field
which want to take help from computational science to detect
unwanted growth of weed along with other crops/plants. Usually in
farming when we farmers grew something due to soil property and
pre available micro seeds additional growth of weeds is there which
spoil the actual outcome of farming as they affect the growth of
planted plants. So weed detection is problem of accurately
identifying the area of weeds so that specific areas can be targeted
for spraying with minimum spraying on the other plants of interest.
PROPOSED WORK
We use YOLOv3 algorithm which is a popular object detection algorithm that uses a deep neural
network to detect and classify objects in images or videos.
Image Acquisition- The recommended approach have been implemented using the weed image
dataset. This approach is used to develop for accurate weed extraction and consists of a number of
phases, including picture capture, edge detection
Image Preprocessing- The input image is pre-processed to prepare it for the detection process.
This typically involves resizing the image to the required input size normalizing the pixel values.
Feature Extraction- The pre-processed image is passed through a deep convolutional neural
network (CNN) to extract features at different scales. YOLOv3 uses a modified version of the
Darknet-53 architecture for feature extraction.
Prediction- The extracted features are then used to make predictions. YOLOv3 predicts bounding
boxes, class probabilities, and confidence scores for multiple objects in a single pass.
Anchor Box Assignment- Anchor boxes are pre-defined bounding boxes of different sizes and
aspect ratios that help predict object locations and sizes. YOLOv3 assigns anchor boxes to the
predicted bounding boxes based on the IoU (Intersection over Union) between the anchor boxes
and ground truth boxes.
PROPOSED WORK
NON maximum separation (NMS)- To remove redundant or overlapping predictions, YOLOv3
applies NMS. It selects the bounding boxes with the highest confidence scores and suppresses
overlapping boxes that have a high IoU threshold .
Output- The final output of the YOLOv3 algorithm consists of the remaining bounding boxes,
their associated class labels, and confidence scores.
Post-processing- Optionally, post-processing steps can be applied to further refine the output.
This may include filtering out low-confidence detections, applying class-specific thresholds, or
performing additional operations based on the specific application.
Visualization- The resulting detections can be visualized by drawing bounding boxes and labels
on the input image.
PROPOSED MODEL
SYSTEM DESIGN
CONCLUSION
In this project, the detection of crop and weeds using YOLOv3 (a machine
learning algorithm) has been done. We have used pyTorch (library-based
machine learning framework), OpenCV, Pandas and other libraries. Accuracy that
have been seen in our datasets is approximately 97% and accuracy depends on
pixels or picture quality and the technology used (i.e. GPU).we used notification
bar for generated results.
REFERENCES
[1] https://www.apriorit.com/dev-blog/599-ai-for-image-processing
[2]https://www.researchgate.net/publication/337464355_OBJECT_DETECTION_AND_IDENOTIFICATION_A_Project_Report
[3]https://pjreddie.com/darknet/yoloo/
[4] https://towardsscience.com/r-cnn-fast-r-cnn-faster-r-cnn-yolo-object-detection-algorithms36d53571365e?gi=951190c93f15
[5]. Alam, M.; Alam, M.S.; Roman, M.; Tufail, M.; Khan, M.U.; Khan, M.T. Constant Machine Learning Based Crop/Weed Detection and
Classification for Variable-Rate Spraying in Agriculture. In Proceedings of the 2020 7th International Conference on Electrical and
Electronics Engineering (ICEEE), Antalya, Turke, 14-16 April 2020.
[6]. Brinkhoff, J.; Vardanega, J.; Robson, A.J. Land Cover Classification of Nine Perennial Crops Using Sentinel-1 and-2 Data. Remote
Sens. 2020.
[7]. Etienne, A.; Saraswat, D. Machine learning approaches to automate weed detection by UAV based sensors. In Autonomous Air
and Ground Sensing Systems for Agricultural Optimization and Phenotyping IV; International Society for Optics and Photonics:
Bellingham, WA, USA, 2019.
[8]. Zhang, S.; Guo, J.; Wang, Z. Combing K-means Clustering and Local Weighted Maximum Discriminant Projections for Weed
Species Recognition. Front. Comput. Sci. 2019.
[9]. Tu, Y.H.; Johansen, K.; Phinn, S.; Robson, A. Measuring canopy structure and condition using multi-spectral UAS imagery in a
horticultural environment. Remote Sens. 2019.
[10]. Bakhshipour, A.; Jafari, A. Evaluation of support vector machine and artificial neural networks in weed detection using shape
features. Comput. Electron. Agric. 2018.
THANK YOU