CSCI598A: Robot Intelligence Jan. 27, 2015 2 3 4 5 6 7 8 Pros/Cons of Global Representations Global representations • Can provide global information (e.g., object poses). • Are NOT invariant to object orientations, image distortions, etc. • Typically need pre-processing steps to locate the objects (e.g., sliding window). 9 Bag-of-words Representation • What is in the picture? 10 Bag-of-words Representation • What is in the picture? 11 Bag-of-words Representation • What is in the picture? 12 Object Bag of ‘words’ 13 Example 1: Bag-of-words models • Orderless document representation: frequencies of words from a dictionary Salton & McGill (1983) US Presidential Speeches Tag Cloud http://chir.ag/phernalia/preztags/ Example 1: Bag-of-words models • Orderless document representation: frequencies of words from a dictionary Salton & McGill (1983) US Presidential Speeches Tag Cloud http://chir.ag/phernalia/preztags/ Example 1: Bag-of-words models • Orderless document representation: frequencies of words from a dictionary Salton & McGill (1983) US Presidential Speeches Tag Cloud http://chir.ag/phernalia/preztags/ Example 1: Bag-of-words models • Orderless document representation: frequencies of words from a dictionary Salton & McGill (1983) Examp2: Texture recognition • Texture is characterized by the repetition of basic elements or textons • For stochastic textures, it is the identity of the textons, not their spatial arrangement, that matters Julesz, 1981; Cula & Dana, 2001; Leung & Malik 2001; Mori, Belongie & Malik, 2001; Schmid 2001; Varma & Zisserman, 2002, 2003; Lazebnik, Schmid & Ponce, 2003 Example 2: Texture recognition histogram Universal texton dictionary • Works pretty well for image-level classification Julesz, 1981; Cula & Dana, 2001; Leung & Malik 2001; Mori, Belongie & Malik, 2001; Schmid 2001; Varma & Zisserman, 2002, 2003; Lazebnik, Ponce, 2003 Csurka et al. (2004), Willamowski et al. (2005), Grauman & Darrell (2005),Schmid Sivic et al.&(2003, 2005) Example 3: Bags of features for scene recognition face, flowers, building • Works pretty well for image-level classification Csurka et al. (2004), Willamowski et al. (2005), Grauman & Darrell (2005), Sivic et al. (2003, 2005) Bag of features (BoF): outline 1. Feature detection 2. Feature description Bag of features (BoF): outline 1. Feature detection 2. Feature description Bag of features (BoF): outline 1. Feature detection 2. Feature description 3. Dictionary learning Codeword ( , , ... ) Codeword ( , , ... ) Codeword ( , , ... ) Bag of features (BoF): outline 1. Feature detection 2. Feature description 3. Dictionary learning 25 26 27 28 29 30 31 32 33 Bag of features (BoF): outline 1. Feature detection 2. Feature description 3. Dictionary learning Bag of features (BoF): outline 1. 2. 3. 4. Feature detection Feature description Dictionary learning Bag-of-features representation