TDP-code example dataset: Flower17 We show experimental settings of three partitions (two tasks, i.e., two scales as inputs for extracting VGG-19 features) on Flower17 datasets. Due to that the extracted deep CNN features are large, we can not upload them through github; authors can extract features by themselves, or contact me, then I will send you the extracted features through other ways.
Running Instructions:
- TDP_main_par1_t1_selftune.m is the main function for conducting Experiments on task 1 of partition 1 on Flower17 datasets.
- after you run all these six functions, you can obtain specific numbers of recognition accuracy.
- for multiple task optimization, the parameters from each single task should be used as the initializing parameters.
If having any problems, please contact: