Image classifier for ppm images.
To find the nearest neighbour for the query, the tool implements a bounded priority queue processing the given images from the training data set. The nearest neighbour class is the guess for the query.
Project carried out in the context of the Data structures and Algorithm course "INFO0902" given by Pr. Geurts, University of Liège.
Grade : 19/20
To determine the best fit for the query, the algorithm used is Dynamic time warping
The training and test sets come from the Google QuickDraw dataset.
Compile the tool
make
To classify one image from testset.txt
, run :
./NNSketch -i index -k k -o ./results trainingset.txt testset.txt
Where :
index
is the index of the image intestset.txt
.k
is the number of images fromtrainingset.txt
to use for the classification.
Example :
./NNSketch -i 1 -k 200 -o ./results trainingset.txt testset.txt
Note : the image selected to be classified is visible in ./results
To test ²the algorithm on all the test set, run :
./NNSketch trainingset.txt testset.txt
The two data sets have the same structure :
number of images
first image
class
number of pixels
pixels
...
last image
- Simon Gardier (Co-author)
- Lei Yang (Co-author)