KyrieRui / YOLO5_River_Reading

Reading river height

This code file is based on Yolo V5, to run this code just follow this bellow steps:

The first version demo:

train with dataset: DATAS2-4

!python train.py --img 640 --batch 16 --epochs 10 --data DATAS2-4/data.yaml --weights yolov5s.pt

test with dataset: DATAS2-4

!python detect.py --source DATAS2-4/valid/images --weights runs/train/exp4/weights/best.pt

Test with dataset: BOPRC_test

!python detect.py --source BOPRC_test --weights runs/train/exp/weights/best.pt

How to make our own dataset:

The website: https://roboflow.com/ , after upload images, use the label tool to label the images, then export the dataset to YOLOv5 format.

After we have the exported yolov5 format dataset, we can use it to train our model same like the code above.

After we trained the model it will locate at 'runs/train/exp/weights/best.pt'

Then we can use it for real time detection:

!python detect.py --source 0 --weights runs/train/exp/weights/best.pt

<<<<<<< Updated upstream

For andorid app, we can use the tflite model, to convert the model to tflite model, just follow this bellow steps:

!python export.py --weights runs/train/exp/weights/best.pt --img 640 --batch 1

=======

For export to tflite model(for embedded into mobile device):

!python export.py --include tflite --imgsz 320 --weights runs/train/exp/weights/best.pt

The version 2 is trained with dataset: DATAS2-4, it have more images which labeled all the numbers from 0 to 9 on the scale