Code of Soft video Parsing method (ASP) on the 9th action (fris-catch) of THUMOS2014 dataset

The codes show an example of our algorithm ASP on the 9th action (fris-catch) of THUMOS2014 dataset. 1.Extract the files from code_and_data_for_ASP. 2.In the dir code, main.m is the start function. In main.m file, 'featureDir' is the dir you place your feature for the dataset. Before you start the main.m function, you should place your feature of dataset under the dir 'stipFeatureThumos'. The feature files of the frisby-catch action are all included in the 'stipFeatureThumos' of the data folder. 3.In data, dir 'parameters' includes some parameters of ASP and SP. 'videoGrounTruth'is the ground truth of THUMOS'14, which includes the ambiguous segments, framerate and total frame number of videos, and the ground truth segments. Dir 'codeAll' and 'confAll' are the BoW coding we've learned. 4.The file apActionBestNew.mat (which will be generated in the results folder) includes the best sub-actions numbers for each action by 3-fold cross validation. 5.The SP method: please add % before the line "featureCodeAll=c_AdjustLabel(iternum,dirIter,featureCodeAll);" in the file 'a_Train_codes.m' which is the added step of local adjustment in ASP method. If you have any question of the codes, please contact miaogen6@gmail.com or xgeng@seu.edu.cn. Thank you for citing our paper: (SP method) Soft Video Parsing by Label Distribution Learning.AAAI 2017 and (ASP method) the paper with the same title which would be published at Frontiers of Computer Science