This project is built using the fastai library to run a deeplearning model that predicts native trees of Kentucky.
- The trained deepleaning model can be found at the following git hub repository https://github.com/chris-kehl/Kentucky_Tree_classifier.
- I used Google Colab to create the machine learning models to take advantage of the GPU's and the convience of cloud computing,
- It is best to view the projects in google colab, the repositories will allow you to open the public Google Colab projects.
To share a little background of the project, basically I know very little about tree identification. I just moved out of the city into a wooded area on on a lake. My home is a new construction, so the builders cut down all of the trees on the property to build. I decided to plant more trees to for the wild life, but I wanted the trees to be native trees that would support the wildlife. Since I was heavily interested in deep learning and wanted to make use of a skill, I decided to make this app.
Now this app is not at all perfect, as a matter of fact most of the time you will get a 50/50 best guess. I have trained over 74 model of trees which are ALLEGHANY SERVICEBERRY', 'AMERICAN BEECH', 'AMERICAN HOLLY', 'AMERICAN HOPHORNBEAM', 'AMERICAN HORNBEAM', 'AMERICAN LINDEN', 'ASH', 'BALD CYPRESS', 'BIGLEAF MAGNOLIA', 'BLACK CHERRY', 'BLACK GUM', 'BLACK LOCUST', 'BLACK MAPLE', 'BLACK OAK', 'BLACK WALNUT', 'BLUE ASH', 'BUR OAK', 'CANADIAN HEMLOCK', 'CAROLINA SILVERBELL', 'CHESTNUT OAK', 'CHINKAPIN OAK', 'COCKSPUR HAWTHORN', 'COMMON WITCHHAZEL', 'CUCUMBERTREE MAGNOLIA', 'DOWNY SERVICEBERRY', 'EASTERN HEMLOCK', 'EASTERN REDBUD', 'EASTERN REDCEDAR', 'EASTERN WHITE PINE', 'FLOWERING DOGWOOD', 'GREEN ASH', 'GREEN HAWTHORN', 'HONEYLOCUST', 'KENTUCKY COFFEETREE', 'LOBLOLLY PINE', 'MOUNTAIN MAPLE', 'MOUNTAIN STEWARTIA', 'NORTHERN CATALPA', 'NORTHERN WHITE CEDAR', 'NORWAY MAPLE', 'OHIO BUCKEYE', 'PAGODA DOGWOOD', 'PAWPAW', 'PECAN', 'PERSIMMON', 'PIGNUT HICKORY', 'PIN OAK', 'PITCH PINE', 'RED BUCKEYE', 'RED MAPLE', 'RED OAK', 'RIVER BIRCH', 'SASSAFRAS', 'SCARLET OAK', 'SHAGBARK HICKORY', 'SHELLBARK HICKORY', 'SHINGLE OAK', 'SHORTLEAF PINE', 'SOURWOOD', 'STRIPED MAPLE', 'SUGAR HACKBERRY', 'SUGAR MAPLE', 'SWEET BIRCH', 'SWEETGUM', 'SYCAMORE', 'TULIP POPLAR', 'UMRRELLA MAGNOLIA', 'VIRGINIA PINE', 'WHITE ASH', 'WHITE FRINGETREE', 'WHITE OAK', 'WILLOW OAK', 'YELLOW BUCKEYE', 'YELLOWWOOD'.
The idea of this app is at the bottom of the page you will upload your photo of a tree you wish to classify. Once uploaded it will give you the number of items you have uploaded. After uploading the picture you will select the classify button and the program will attempt to give the tree that the model predicts your photo is and a probability. I have ranged from a 90% accuarcy to 20%. Don't expect precision results with this application, since it's just a getting familiar with image recognition using the fastai introductory library. To get more accurate I plan to research more image recognition algorithms, while finding more labled data for native trees of Kentucky.