Open tymanman opened 3 years ago
The same question, I think this comparision may not be reasonable
Hi, In the Table 2 of the paper, you compare the average ap of mutiple runs of 10 seeds with the results given by TFA which is a average one of 30 seeds.However it is obvious that the former is anyway better than the latter according to the graph shown in TFA, So maybe this seems to be a little unresonable? Look forward to your reply.
Your observation is right, it should be 30 seeds here. I'd rather believe this is a typo in the paper, but I didn't get same result as in Table2 yet.
Hi, In the Table 2 of the paper, you compare the average ap of mutiple runs of 10 seeds with the results given by TFA which is a average one of 30 seeds.However it is obvious that the former is anyway better than the latter according to the graph shown in TFA, So maybe this seems to be a little unresonable? Look forward to your reply.
Your observation is right, it should be 30 seeds here. I'd rather believe this is a typo in the paper, but I didn't get same result as in Table2 yet.
I notice that the author makes a tricky test with augmentation TEST.AUG.ENABLED=True
while other competing methods did not do so. After removing this, I obtained the results as shown below, which are somewhat lower than the reported numbers. How about your reproduced numbers?
split1 | split2 | split3 | |||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1shot | 2shot | 3shot | 5shot | 10shot | 1shot | 2shot | 3shot | 5shot | 10shot | 1shot | 2shot | 3shot | 5shot | 10shot | |
AP50 | 69.10 | 68.33 | 68.11 | 70.42 | 72.93 | 65.03 | 65.75 | 67.33 | 70.14 | 71.03 | 69.02 | 70.18 | 68.29 | 72.97 | 72.66 |
bAP50 | 78.44 | 77.13 | 74.29 | 74.67 | 76.74 | 79.47 | 78.06 | 75.48 | 78.52 | 77.85 | 79.58 | 78.36 | 75.66 | 78.86 | 77.42 |
nAP50 | 41.08 | 41.92 | 49.57 | 57.68 | 61.51 | 21.69 | 28.80 | 42.87 | 45.00 | 50.58 | 37.31 | 45.65 | 46.19 | 55.31 | 58.36 |
Hi, In the Table 2 of the paper, you compare the average ap of mutiple runs of 10 seeds with the results given by TFA which is a average one of 30 seeds.However it is obvious that the former is anyway better than the latter according to the graph shown in TFA, So maybe this seems to be a little unresonable? Look forward to your reply.