You are on page 1of 6

Landmarks

Recognition
System Using AI
Problem
 Statement
Landmarks recognition system is one of the important part in computer vision but also the most difficult
 This concept is used in the AR technology but we have yet to achieve a good AI model
 The sheer amount of landmarks just for a city is overwhelming like we if are classifying a landmarks we
need 100 or so image and if we have 100 places then we need to use data of 100*100=10000 image

 The prediction output of algorithm like still and random forert is still very low
INTRODUCTION
 The problem comes from a famous Kaggle competition, the Google Landmark
Recognition Challenge. Training set contains over 1.2 million images spread
across 14,951 classes of landmarks, varying from one to thousands of images per
class. This problem of extreme classification is something that is very prevalent in
the data science community today with the advancement of deep learning. One
important thing to note about this dataset is that the test set given included many
images that were not landmarks at all, that we refer to as “junk” images.
PURPOSED SYSTEM
GATHERING DATA
The first step to our analysis is to mind the images of landmarks and store in
the data
POST DATA AND PRE PROCESSING
This method included things he VGG16 pre-trained model on the data along
with delf for dealing with difficult text image
MODEL TRAINING
we used the VGG16 pre-trained on Google ImageNet dataset to predict the
landmarks. We observed that ImageNet weights are successfully able to capture
the generic features from our images

 FINAL MODEL
 Our model trained with VGG16 architecture could give us an accuracy of 78.60%
on validation set, we worked exhaustively on increasing accuracy looking at the
effects of varying parameters before moving onto other models like Inception
and Resnet
OBJECTIVE
To improve the accuracy of model used in computer vision in landmarks
To reduce the no of images required to recognize the landmarks the city

You might also like