💡 Train Weed Detection Model on Weed Dataset

[qiugai]

Start Date

3/7/23

Implementation PR

No response

Reference Issues

No response

Summary

We need to change from our generic yolov5 general object dataset, to a custom model using a dataset of weeds on sidewalks and pavements.

Basic Example

Detect weeds not fridges and people and chairs

Drawbacks

Training take long time :(

[qiugai]

Data is being trained on Roboflow, ~1000 images, 50% | 30% | 20% | Train | Valid | Test split

[ZM009]

Experienced the results of the training yesterday. Holy moly

[ZM009]

Data training worked fine, model had some different results when training more.