annotated flower / plant image dataset

Question

I'd like to try and build a classification model for plants and flowers pictures, similar to what Daniel Nouri did with his app: http://danielnouri.org/notes/2014/09/13/identifying-birds,-butterflies,-and-wildflowers-with-a-snap/

For this I'd need an annotated set of plant and/or flowers (I don't have a definite goal) to start training with. I tried contacting the author, to no avail and I haven't been able to find anything interesting.

Kaggle released a dataset of leaves "shadows": https://www.kaggle.com/c/leaf-classification which can be interesting to work with, but makes the transition from raw picture to classification hard.

Any idea where I could find such a dataset?

Answer 1

I finally found exactly what I was looking for : 113,205 pictures of 1000 different plants. It was curated and is hosted as part of the CLEF competitions held every year, and it is therefore enriched every year through crowdsourcing.

http://www.imageclef.org/lifeclef/2016/plant

Answer 2

I understand you need a big amount of photographs of plants with known species name. A place to start is plants photographs in Wikimedia Commons, with lots of images of plants, and for very common species up to hundreds of photographs of each specie. The only drawback I see is that species names is not often in a machine readable form and you will need to do some work to anotate images. To do this, Wikidata might help to relate each species (in Wikidata) with one category in Commons.