This project was designed for the purpose of from a list of plants with their common and scientific names, to download a sampleset of images for each plant in the list and train a CNN for the purpose of identifying those plants.
Note: run all project files with Python 3
The main entrypoint to the webscrapping app is through driver.py. Running python driver.py will open a predetermined list of plant names in names.csv the format "Common Name,Scientific name". It will then go through four different web sources of plant imagery (one of which being google images) and save image links to a file called links.csv. This links can then be downloaded by running python downloader.py. Within downloader.py you can set the size variable to the desired size in pixles that all the downloaded images will be sized to (the majority of image classification algorithms will require equally sized images).
In the collect_names directory, a list of plants native to a zipcode can be obtained by runing python collect_names.py. The results will be stored in names.csv which can be used for the scraper. Modify the zipcode variable in collect_names.py to collect names from your desired zipcode.
The main entrypoint into the image classification app is through tfscript.py. Running this will already assume you have results stored in the output directory from the downloader script. Running python tfscript.py will pass the images through a convolutional neural network and then test the network on a random subset of images and report accuracy.