billmails / apple-harvesting-robot-vision-system

creating a vision system for apple harvesting robot using YOLO algorithm

vision system for apple harvesting robot YOLOv4

creating a vision system for apple harvesting robot using YOLO algorithm

steps to develop a custom detector

1.Create yolov4 and training folders in your google drive

received gpu

2.Mount drive, link your folder and navigate to the yolov4 folder

3.Clone the Darknet git repository

4.Create & upload the files we need for training ( i.e. “obj.zip” , “yolov4- custom.cfg”, “obj.data”, “obj.names” and “process.py” ) to your drive

Data set construction

Single object with no occlusion, Multiple objects with occlusion, Clusters of apples, Illumination variation, Shading conditions, Multiple objects with or without occlusion

5.Make changes in the Makefile to enable OPENCV and GPU

6.Run make command to build darknet

7.Copy the files “obj.zip”, “yolov4-custom.cfg”, “obj.data”, “obj.names”, and “process.py” from the yolov4 folder to the darknet directory

8.Run the process.py python script to create the train.txt & test.txt files

9.Download the pre-trained YOLOv4 weights

10.Train the detector

11.Check performance

performence graph

performence check by mean average precision(mAP)

12.Test your custom Object Detector

Test results on images

Test results on webcam images

Testing on a higher illumination condtion Testing on a highershading condtion

Test results on videos

medium1.mp4

medium2.mp4

medium3.mp4