fabianrathke / octsegmentation Goto Github PK

Currently the c-implementation of sum-product only returns singleton terms of $q_c$. When calculating the terms of the objective function, a slow implementation of Matlab is required that explicitly relies on the $\Omega$-Terms. Adapting the c-function could speed up this calculation significantly.

Compile "compileMex.m"
Prompt the following error:
Cannot open include file: “stdlib.h”: No such file or directory

I run into a bug when trying to run the demo " predSrinivasan.m" on the Srinivasan dataset:

predSrinivasan
******** FILE 1 ********
********* SCAN 5 *********
No cached appearance terms found, calculating them
Error using checkPredictionGlobal (line 4)
Not enough input arguments provided

Error in fetchPatches (line 26)
checkPredictionGlobal(filename);

Error in collectTestData (line 43)
testData = fetchPatches(file.name,idxSet,options);

Error in predAppearance (line 64)
a = tic; patches = collector.name(files(i),collector.options); timeFetchingPatches = toc(a);

Error in initQC (line 109)
predictionA = predAppearance(files(file),collector,models.appearanceModel,options);

Error in predVariational (line 84)
initQC;

Error in predSrinivasan (line 60)
predictionTmp = testFunc.name(testFile,collectorTest,struct(),model,testFunc.options);

I tried to fix the problem by changing "checkPredictionGlobal(filename)" to "checkPredictionGlobal(filename, options.labelID)" but matlab crashes after the modification.

GPUmat not supported anymore, Matlab developed good GPU support with even more functions implemented for use on GPU.