Plant leaf disease detection

I want to develop a model for detection of disease looking at plant leaves and stem.
I am not sure if it is classification problem or object detection problem. I tried classification but it fails in the real life. Some issues that need to be addressed are:

  1. The disease spots are very small compared to entire image size.
  2. Shape of disease spots is irregular and the patterns are based on color spots and color gradient.
  3. Background of leaf can sometime look like a disease spot.

Can anyone suggest the approach for this, i would be very grateful. Thanks

1- Train with more data. See kaggles competions like this.
2- Small details are harder to pick. So, you may want to crop the image around leaf. You could have an object detection to detect the leaf -> crop the image around it -> classify

Thanks Victor, this make sense I will try it.