- Author: Chiyuan Cheng (cyuancheng AT gmail DOT com)
- Last updated: Aug 29, 2019
To create an AI detector for common lawn weeds using computer vision and deep learning. The 7 common lawn weeds are:
- Annual Bluegrass
- Crabgrass
- Creeping Charlie
- Dandelion
- Dollar weed
- Oxalis
- White clover
The training and test imaging dataset was collected from Google image and stored on google drive for analysis.
- Image augmentation was used to improve the performance of deep learning in imaging classification.
- Dataset contains 1830 training data and 457 testing data.
- resnet50 was used for the deep learning model.
- 95% accuracy can be achieved to predict the lawn weed.
- The code for this project can be found here.
The credit was given fast.ai courses for the fast.ai package.