Implementation of SVM active proposed here.
Needs keras
, PIL
, pandas
and a few other common packages.
First, have a look at the config.yaml
file.
Execute python get_data_ready.py
to get the data ready. This will download
tiny-imagenet-200.zip
, unzip it into a folder and create an npy
file
of the loaded feature vectors obtained by using keras
.
Then execute python run_svm_with_feedback.py
. This will create a plot
using the configuration specified in config.yaml
which shows the performance
of the SVM when samples to get feedback for are chosen using the algorithm
in the paper as opposed to when they are chosen randomly.
The SVM which is used in this implementation is the one used by SMQTK. The path is specified in config.yaml
.