local M = { } -- http://stackoverflow.com/questions/6380820/get-containing-path-of-lua-file function script_path() local str = debug.getinfo(2, "S").source:sub(2) return str:match("(.*/)") end function M.parse(arg) local cmd = torch.CmdLine() cmd:text() cmd:text('OpenFace') cmd:text() cmd:text('Options:') ------------ General options -------------------- cmd:option('-cache', paths.concat(script_path(), 'work'), 'Directory to cache experiments and data.') cmd:option('-save', '', 'Directory to save experiment.') cmd:option('-data', paths.concat(os.getenv('HOME'), 'openface', 'data', 'casia-facescrub', 'dlib-affine-sz:96'), -- 'dlib-affine-224-split'), 'Home of dataset. Images separated by identity.') cmd:option('-manualSeed', 2, 'Manually set RNG seed') cmd:option('-cuda', true, 'Use cuda.') cmd:option('-device', 1, 'Cuda device to use.') cmd:option('-cudnn', true, 'Convert the model to cudnn.') ------------- Data options ------------------------ cmd:option('-nDonkeys', 2, 'number of donkeys to initialize (data loading threads)') ------------- Training options -------------------- cmd:option('-nEpochs', 1000, 'Number of total epochs to run') cmd:option('-epochSize', 250, 'Number of batches per epoch') cmd:option('-epochNumber', 1, 'Manual epoch number (useful on restarts)') -- GPU memory usage depends on peoplePerBatch and imagesPerPerson. cmd:option('-peoplePerBatch', 15, 'Number of people to sample in each mini-batch.') cmd:option('-imagesPerPerson', 20, 'Number of images to sample per person in each mini-batch.') cmd:option('-testing', true, 'Test with the LFW.') cmd:option('-testBatchSize', 800, 'Batch size for testing.') cmd:option('-lfwDir', '../data/lfw/aligned', 'LFW aligned image directory for testing.') ---------- Model options ---------------------------------- cmd:option('-retrain', 'none', 'provide path to model to retrain with') cmd:option('-modelDef', '../models/openface/nn4.def.lua', 'path to model definiton') cmd:option('-imgDim', 96, 'Image dimension. nn2=224, nn4=96') cmd:option('-embSize', 128, 'size of embedding from model') cmd:option('-alpha', 0.2, 'margin in TripletLoss') cmd:text() local opt = cmd:parse(arg or {}) os.execute('mkdir -p ' .. opt.cache) if opt.save == '' then opt.save = paths.concat(opt.cache, os.date("%Y-%m-%d_%H-%M-%S")) end os.execute('mkdir -p ' .. opt.save) return opt end return M