Leaves as organs available most of the time of growing season represent important element in the process of recognition an unknown plant. We present the process of automatic leave´s recognition on these web pages - make public information about the project, links to used datasets etc. You can also try algorithms on your own collected leaves. Please give your attention to a recommendation about collecting and digitizing unknown leaves.
GAUK project no. 2012/524512 - Computer-aided woody plants identification based on leaf analysis techniques outcome.
To achieve the best results please try to keep some rules. It is very important to respect that the application approaches distinctly to a definition of leaf in comparison with standard botanical terminology. By compound leaves the application only processes separate leaflets and in case you upload a whole leaf (f. e. red ash Fraxinus excelsior, buckeye Aesculus hippocastanum etc.), you will get totally wrong results. By palmated leaves use middle leaflet, by pinnated leaves the leafletsleaves from the middle pairs. This approach makes mathematical operations such as differentiation of petiol and blades easier , or at the same time a total description of shape; affirmative is as well a benefit for the real applicability to laymen who don´t distinguish by some leaves if they are compound or not. Its part play as well size of leaf which don´t fit at some compound leaves to size A3 (f.e. Gymnocladus dioicus - Kentucky coffeetree).
You can take a picture of leaf f.e. laid on sheet but it´s better to scan it with the white background on 300DPI resolution - by this way are also obtained items of the self databases. The application at this stadium of development doesn´t allowed a ordinarily picture segmentation and you have to realize this action manually before uploading a picture eventually.
Image is in process, please wait. Total processing time is based on the size of recorded picture and on the speed of your connection. The very process on server takes maximally thirty seconds.