top of page
B.U. Solid State Laboratory, May 2013 - July 2014
B.U. Solid State Laboratory, May 2013 - July 2014
B.U. Solid State Laboratory, May 2013 - July 2014
Spatial Pyramid Matching for Scene Classification
CMU16-720, Computer Vision
September 2020
Abstract:
Given a set of images, the goal was to determine the location of the scenes using Spatial Pyramid Matching. This representation is based off of the bag of visual words approach.
![](https://static.wixstatic.com/media/94d43b_0a3672d0f34a4cf78666f5cf2c38acdd~mv2.jpg/v1/fill/w_385,h_287,al_c,q_80,usm_0.66_1.00_0.01,enc_auto/94d43b_0a3672d0f34a4cf78666f5cf2c38acdd~mv2.jpg)
Process:
I started by applying a filter bank to each our images to tease out the high frequency signals. The filters consisted of variations of the Gaussian and Gaussian-Laplace filters. From here, I took samples of pixels from the filter responses and passed it to a K-mean clustering algorithm to generate a 'visual words' dictionary. This was done over thousands of training images to generate a model where I could create word maps describing each scene image. Finally I developed a recognition model using the spatial pyramid matching technique which mapped each test image to the closest historgram describing the image.
![](https://static.wixstatic.com/media/94d43b_ed6dd761e62a4360af9618ade29ad5e8~mv2.png/v1/fill/w_449,h_661,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/94d43b_ed6dd761e62a4360af9618ade29ad5e8~mv2.png)
![](https://static.wixstatic.com/media/94d43b_5363cf39b9ed4356ba59fb8c96c3d23c~mv2.png/v1/fill/w_469,h_266,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/94d43b_5363cf39b9ed4356ba59fb8c96c3d23c~mv2.png)
Results:
To the right are some examples of the word maps generated using this model. I was able to achieve an accuracy of 65.5% using this image classification model.
bottom of page