Hi Hivers,
When I was overwhelmed by the chaos of the city, I finally found a place in nature where I could relax. I have been stockpiling leaf photographs for some time now. I have been thinking about a computer vision project related to this for a long time. I'm going to use the leaf photos as training data to develop an artificial intelligence model. For now, I don't have a graphics card that can do this, but I know that one day I will have a graphics card that can do this. If you are familiar with computer vision projects, you can imagine that the hard part is collecting the data.
I put here the Latin names of the plants I photographed, there is tons of information on the internet for all of them. To train the dataset, you may need hundreds or even thousands of photos for a species. There are a lot of open data bases for plants on the internet, but I think it would be a more sensible choice to focus on the species that are in my environment.Trying to create and use a training model for plants not found in the local environment would be a waste of time and unnecessary GPU usage. GPU prices have been high for many years due to cryptocurrency miners. I think the most logical choice is to rent a server with GPU support and train the model there.
The algorithms used in computer vision projects are generally successful for objects that differ in shape. Computer vision is even mentioned as one of the most successful subdisciplines of artificial intelligence. But I guess it would be a bit more laborious to do such a work for plants. Because leaves are objects that are similar in shape. It is not easy to tell which species a plant belongs to, at least from a distance, even with the human eye. I still want to do a model training for it.
I will continue to collect photos of plants for training data and will share my progress as I have time.
Have a nice weekends
For the best experience view this post on Liketu