Hi, Can you please clarify the experimental setup?
In your paper, it mentions that 'For the comparison with the previous approaches, we used both single and interacting hand (SH+IH) images of the H+M split. For ablation studies, we used SH+IH images of the H split. And in table 5, it provide the comparision in both single and interacting hand (SH+IH) images of the H+M split :
But when envaluated in SH+IH and H+M split , the model get the results with given checkpoint like here:
When envlauted in SH+IH and only H split by uncomment the code here , the model get the reults like here :
It seems that the latter is closer to the results mentioned in the article. But it does not use M split.
Hi, Can you please clarify the experimental setup? In your paper, it mentions that 'For the comparison with the previous approaches, we used both single and interacting hand (SH+IH) images of the H+M split. For ablation studies, we used SH+IH images of the H split. And in table 5, it provide the comparision in both single and interacting hand (SH+IH) images of the H+M split : But when envaluated in SH+IH and H+M split , the model get the results with given checkpoint like here:
When envlauted in SH+IH and only H split by uncomment the code here , the model get the reults like here : It seems that the latter is closer to the results mentioned in the article. But it does not use M split.
Looking forward your reply! Thank you.