Differences

This shows you the differences between two versions of the page.

--- ece4580:module_classification [2017/01/29 16:11] – pvela
+++ ece4580:module_classification [2024/08/20 21:38] (current) – external edit 127.0.0.1
@@ Line 48: / Line 48: @@
 */
+__**Week #1: Clustering to Define Words**__ \\
+  - Study [[https://sites.google.com/site/dataclusteringalgorithms/k-means-clustering-algorithm|k-means clustering algorithm]] and the algorithmic steps for k-means clustering.
+  - Download (or clone) the clustering skeleton code [[https://github.gatech.edu/fchu9/ECE4580_LearningModules/tree/master/objectRecognition/clustering|here]]
+  - Implement k-means clustering algorithm working in RGB space by following the algorithmic steps. You are welcome to implement from scratch without skeleton code.
+  - Test your algorithm on segmenting the image //segmentation.jpg// using k=3
+  - Try different random initialization and show corresponding results.
+  - Comment on your different segmentation results.
+//Matlab Notes:// Matlab has several functions that can assist with the calculations so that you do not have to process the data in a for loops.  For example, there is **pdist2**, the mean function can process data row-wise or column-wise if specified properly, and there are ways to perform sub-assignments using either a binary array or the list of indices themselves.  You should be taking advantage of these opportunities to have compact code.
+__**Week #2: Object Recognition**__\\
+  - Study the [[https://www.youtube.com/watch?v=unCLMMsZAh8|bag-of-words]] approach for classification/Recognition task
+  - We begin with implementing a simple but powerful recognition system to classify //faces// and //cars//.
+  - Check [[https://github.gatech.edu/fchu9/ECE4580_LearningModules/tree/master/objectRecognition/simpleBoW|here]] for skeleton code. First, follow the README to setup the dataset and vlfeat library.
+  - In our implementation, you will find [[http://www.vlfeat.org/index.html|vlfeat library]] very useful. One may use vl_sift, vl_kmeans and vl_kdtreebuild.
+  - Now, use first 40 images in both categories for training.
+  - Extract SIFT features from each image
+  - Derive k codewords with k-means clustering in module 1.
+  - Compute histogram of codewords using [[https://en.wikipedia.org/wiki/K-d_tree|kd-tree algorithm]] using vlfeat.
+  - Use the rest of 50 images in both categories to test your implementation.
+  - Report the accuracy and computation time with different k
+__**Week #3: Spatial Pyramid Matching (SPM)**__\\
+Usually objects have different properties across the spatial scales, even though they may appear common at one given scale.  Consequently, differentiation of objects is often improved by concatenating feature vectors or agglomerating feature descriptors that exist at different scales. This set of activities will explore how well that can work.
+  - Study [[https://hal.archives-ouvertes.fr/file/index/docid/548585/filename/cvpr06_lana.pdf|Spatial Pyramid Matching]] which can improve BoW apporach by concatenating histogram vectors.
+  - We will implement a simplified version of SPM based on your molude 2
+  - First, for each traning image, divide it equally into a 2 × 2 spatial bin.
+  - Second, for each of the 4 bins, extract the SIFT features and compute the histograms of codewords as in module 2
+  - Third, concatenate the 4 histogram vectors in a fixed order. (hint: the a vector has 4k dimension.)
+  - Forth, concatenate the vector you have in module 2 with this vector (both weighted by 0.5 before concatenated).
+  - Finally, use this 5k representation and re-run the training and testing again.
+  - Compare the results from module 3 and module 2. Explain what you observe.
 -----------------