Mercurial > cortex

comparison org/literature-review.org @ 377:80cd096682b2

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
reviewing ullman's stuff.
author Robert McIntyre <rlm@mit.edu>
date Thu, 11 Apr 2013 06:19:59 +0000
parents 057d47fc4789
children 8e62bf52be59
comparison
equal deleted inserted replaced
376:057d47fc4789 377:80cd096682b2
206 (true-positives/total-in-class) 206 (true-positives/total-in-class)
207 207
208 cross-validation = train the model on two different sets to prevent 208 cross-validation = train the model on two different sets to prevent
209 overfitting. 209 overfitting.
210 210
211 nifty, relevant, realistic ideas
212 He doesn't confine himself to unplasaubile assumptions
211 213
212 214
213 215
214 216
215 ** Getting around the dumb "fixed training set" methods 217 ** Getting around the dumb "fixed training set" methods
237 range of training samples, since his model changes complexity 239 range of training samples, since his model changes complexity
238 depending on the number of training samples. The simpler models do 240 depending on the number of training samples. The simpler models do
239 better with few training points, and the more complex ones do 241 better with few training points, and the more complex ones do
240 better with many training points. 242 better with many training points.
241 243
244 The final model had intermediate complexity between published
245 extremes.
246
242 The more complex models must be able to be initialized efficiently 247 The more complex models must be able to be initialized efficiently
243 from the less complex models which they replace! 248 from the less complex models which they replace!
244 249
245 250
246 ** Non Parametric Models 251 ** Non Parametric Models
247 252
248 *** Visual features of intermediate complexity and their use in classification 253 *** 2002 Visual features of intermediate complexity and their use in classification
249 254
250 *** The chains model for detecting parts by their context 255
256
257 *** 2010 The chains model for detecting parts by their context
251 258
252 Like the constelation method for rigid objects, but extended to 259 Like the constelation method for rigid objects, but extended to
253 non-rigid objects as well. 260 non-rigid objects as well.
254 261
255 Allows you to build a hand detector from a face detector. This is 262 Allows you to build a hand detector from a face detector. This is
256 usefull because hands might be only a few pixels, and very 263 usefull because hands might be only a few pixels, and very
257 ambiguous in an image, but if you are expecting them at the end of 264 ambiguous in an image, but if you are expecting them at the end of
258 an arm, then they become easier to find. 265 an arm, then they become easier to find.
259 266
260 267 They make chains by using spatial proximity of features. That way,
268 a hand can be idntified by chaining back from the head. If there
269 is a good chain to the head, then it is more likely that there is
270 a hand than if there isn't. Since there is some give in the
271 proximity detection, the system can accomodate new poses that it
272 has never seen before.
273
274 Does not use any motion information.
275
276 *** 2005 A Hierarchical Non-Parametric Method for Capturing Non-Rigid Deformations
277
278 (relative dynamic programming [RDP])
279
280 Goal is to match images, as in SIFT, but this time the images can
281 be subject to non rigid transformations. They do this by finding
282 small patches that look the same, then building up bigger
283 patches. They get a tree of patches that describes each image, and
284 find the edit distance between each tree. Editing operations
285 involve a coherent shift of features, so they can accomodate local
286 shifts of patches in any direction. They get some cool results
287 over just straight correlation. Basically, they made an image
288 comparor that is resistant to multiple independent deformations.
289
290 !important small regions are treated the same as nonimportant
291 small regions
292
293 !no conception of shape
294
295 quote:
296 The dynamic programming procedure looks for an optimal
297 transformation that aligns the patches of both images. This
298 transformation is not a global transformation, but a composition
299 of many local transformations of sub-patches at various sizes,
300 performed one on top of the other.
301
302 *** 2006 Satellite Features for the Classification of Visually Similar Classes
303
304 Finds features that can distinguish subclasses of a class, by
305 first finding a rigid set of anghor features that are common to
306 both subclasses, then finding distinguishing features relative to
307 those subfeatures. They keep things rigid because the satellite
308 features don't have much information in and of themselves, and are
309 only informative relative to other features.
310
311 *** 2005 Learning a novel class from a single example by cross-generalization.
312
313 Let's you use a vast visual experience to generate a classifier
314 for a novel class by generating synthetic examples by replaceing
315 features from the single example with features from similiar
316 classes.
317
318 quote: feature F is likely to be useful for class C if a similar
319 feature F proved effective for a similar class C in the past.
320
321 Allows you to trasfer the "gestalt" of a similiar class to a new
322 class, by adapting all the features of the learned class that have
323 correspondance to the new class.