We are trying to reproduce the results by using the codes of neuraltalk2. We have obtained poor results on flicker8k. After adjusting the parameters, we still cannot get results that are comparable to the ones reported in other papers. Can anyone give us some suggestions on the experiments?
We are trying to reproduce the results by using the codes of neuraltalk2. We have obtained poor results on flicker8k. After adjusting the parameters, we still cannot get results that are comparable to the ones reported in other papers. Can anyone give us some suggestions on the experiments?