Why Recall@K is equal to zero

[gobigrassland]

I try to execute the command " python hdc_test.py -d cars196 -c HDC" using the model trained by you. But
in the script hdc_test.py, when choose HDC method, the function test_cars.eval is called. So it occurs an error. i think test_cars.eval_HDC should be called. I modify the error. In the function eval_HDC, load_lmdb should be given to feature_dim/3(feature_dim=384). After i modify all the errors, the computed recall@K is equal to zero, no matter what's the value of K. I do not know why?

---

stanford cars mean recall@ 1 : 0.000000
stanford cars mean recall@ 2 : 0.000000
stanford cars mean recall@ 4 : 0.000000
stanford cars mean recall@ 8 : 0.000000
stanford cars mean recall@ 16 : 0.000000
stanford cars mean recall@ 32 : 0.000000

[gobigrassland]

I am very confused that labels of query images are indices of list. In the script test_cars196.py, i get labels of query images from test_8131.txt instead of test_label.dat. Then the results of recall@K seem normal.
Using the model author trained, Recall @k on CARS196 test set as follow

  | Recall@K(%)
  | 1 | 2 | 4 | 8 | 16 | 32
HDC+Contrastive-1-128 | 48.1 | 60.9 | 72.4 | 82.1 | 89.7 | 94.4
HDC+Contrastive-2-128 | 59.0 | 71.3 | 81.0 | 87.7 | 93.0 | 96.2
HDC+Contrastive-3-128 | 65.9 | 77.1 | 85.9 | 92.0 | 95.6 | 97.7
HDC+Contrastive-384 | 72.1 | 82.3 | 88.9 | 93.5 | 96.6 | 98.3

there is little difference.