files [options] options: --rootdir, -d root directory of files descriptors are compute from [optional, default is '.'] --namefilters, -t name filters for files to be listed, e.g. "*.png" "*.jpg" [required] --filelist, -f file that contains existing list of filenames [optional, if not provided all matching files in and below rootdir are listed] --outputfile, -o output filelist filename [optional, if not provided, output is console.] --random-sample, -r random shuffle and truncate file list to given size [optional] --seed, -s seed value for random-sampling [optional, defaultis current time]
compute_descriptors
对由generate_filelist产生的索引中的图片进行特征提取
descriptors对应的generator有4种：
galif
gist
shog
tinyimage
用例：
1
compute_descriptors.exe compute galif -r . -f test -o galif_
1 2 3 4 5 6 7
compute <generator> [options] options: --rootdir, -r root directory of data descriptors are computed from [required] --filelist, -f file that contains filenames of data (images/models) [required] --output, -o output prefix [required] --parameters, -p parameters for generator construction [optional] (default: params defined in generator) --numthreads, -t number of threads for parallel computation [optional] (default: number of processors)
compute_vocabulary [options] options: --descfile, -d descriptors file [required] --sizefile, -s file that contains number of words per descriptor[optional] 此处若设置，则-n也必须设置 --numsamples, -n number of words randomly extracted from descriptor file [optional, but sizefile must also be specified] 此处若设置，则-s也必须设置 --numclusters, -c number of clusters/visual words to generate [required] --outputfile, -o output file [required] --numthreads, -t number of threads for parallel computation (default: number of processors) [optional] --maxiter, -i kmeans stopping criterion: maximum number of iterations (default: 20) [optional] --minchangesfraction, -m kmeans stopping criterion: number of changes (fraction of total samples) (default: 0.01) [optional]
compute_histvw [options] options: --vocabulary, -v filename of the vocabulary to be used for quantization [required] --descriptors, -d filename of the descriptors to convert into histograms of visual words [required] --positions, -p positions data for features [required] --output, -o filename of the output file of histograms of visual words [required] --quantization, -q quantization method {hard,fuzzy} [required] --sigma, -s sigma for gaussian weighting in fuzzy quantization [required (with 'fuzzy' quantization only)] --pyramidlevels, -l number of spatial pyramid levels [optional, default 1]
compute_index [options] options: --histvw, -h filename to vector of histograms of visual words [required] --output, -o filename of the output index file [required] --tfidf, -t two strings specifying tf and idf function to be used (eg. -t constant constant) [required]
image_search [options] options: --queryimage, -q filename of image to be used as the query [required] --searchptree, -s filename of the JSON file containing parameters for the search manager [optional, if not provided, --searchparams must be given] --searchparams, -m parameters for the search manager [optional, if not provided, --searchptree must be given] --vocabulary, -v filename of vocabulary used for quantization [optional, only required with bag-of-features search] --filelist, -l filename of images filelist [required] --generatorptree, -p filename of the JSON file containing generator name and parameters [optional, if not provided, generator's default values are used'] --numresults, -n number of results to search for [optional, if not provided all distances get computed] --generatorname, -g name of generator [optional, if given, we will use generator's default parameters and ignore --generatorptree]