Validation confusion

[zx-explorer]

Clearly, the pre-change image results are divided into 7 categories, resulting in 49 possible changes. Why, then, is the parameter passed into SCDD_eval_all set to 37?

Traceback (most recent call last): File "/home/dmx_bs/MambaCD2/MambaCD/changedetection/script/train_MambaSCD.py", line 259, in <module> main() File "/home/dmx_bs/MambaCD2/MambaCD/changedetection/script/train_MambaSCD.py", line 255, in main trainer.training() File "/home/dmx_bs/MambaCD2/MambaCD/changedetection/script/train_MambaSCD.py", line 145, in training kappa_n0, Fscd, IoU_mean, Sek, oa = self.validation() ^^^^^^^^^^^^^^^^^ File "/home/dmx_bs/MambaCD2/MambaCD/changedetection/script/train_MambaSCD.py", line 203, in validation kappa_n0, Fscd, IoU_mean, Sek = SCDD_eval_all(preds_all, labels_all, 37) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/dmx_bs/MambaCD/changedetection/utils_func/mcd_utils.py", line 209, in SCDD_eval_all assert unique_set.issubset(set([x for x in range(num_class)])), f"unrecogniz ed label number, {unique_set}, {set([x for x in range(num_class)])}" ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AssertionError: unrecognized label number, {0, 8, 9, 10, 11, 12, 14, 15, 16, 17, 18, 20, 21, 23, 24, 26, 27, 32, 33, 35, 36, -4, -3}, {0, 1, 2, 3, 4, 5, 6, 7, 8 , 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36}

[ChenHongruixuan]

Hi,

Thank you for yoru question. The SECOND dataset has six semantic categories and one unchanged category. Therefore, the final category number is 6x6 + 1 = 37. Please take a look at the introduction of this dataset: https://captain-whu.github.io/SCD/

Best,

[zx-explorer]

In the raw paper of SECOND, the author claimed that 30 categories and one unchanged class are considered. Why 37 in total?

[ChenHongruixuan]

grount truth has 30 categories, but there are cases of misclassification, so 37 categories must be considered for accuracy assessment.

[zx-explorer]

Thank you for your patient answer. So why the assertion in SCDD_eval_all can be always satisfied? The preds can be negative according to the calculation of preds_csd in the file "train_MambaSCD.py".
`def SCDD_eval_all(preds, labels, num_class):
hist = np.zeros((num_class, num_class))
for pred, label in zip(preds, labels):
infer_array = np.array(pred)
unique_set = set(np.unique(infer_array))
assert unique_set.issubset(set([x for x in range(num_class)])), "unrecognized label number"
label_array = np.array(label)
assert infer_array.shape == label_array.shape, "The size of prediction and target must be the same"
hist += get_hist(infer_array, label_array, num_class)

hist_fg = hist[1:, 1:]
c2hist = np.zeros((2, 2))
c2hist[0][0] = hist[0][0]
c2hist[0][1] = hist.sum(1)[0] - hist[0][0]
c2hist[1][0] = hist.sum(0)[0] - hist[0][0]
c2hist[1][1] = hist_fg.sum()
hist_n0 = hist.copy()
hist_n0[0][0] = 0
kappa_n0 = cal_kappa(hist_n0)
iu = np.diag(c2hist) / (c2hist.sum(1) + c2hist.sum(0) - np.diag(c2hist))
IoU_fg = iu[1]
IoU_mean = (iu[0] + iu[1]) / 2
Sek = (kappa_n0 * math.exp(IoU_fg)) / math.e

pixel_sum = hist.sum()
change_pred_sum  = pixel_sum - hist.sum(1)[0].sum()
change_label_sum = pixel_sum - hist.sum(0)[0].sum()
change_ratio = change_label_sum/pixel_sum
SC_TP = np.diag(hist[1:, 1:]).sum()
SC_Precision = SC_TP/change_pred_sum
SC_Recall = SC_TP/change_label_sum
Fscd = stats.hmean([SC_Precision, SC_Recall])
return kappa_n0, Fscd, IoU_mean, Sek`

[ChenHongruixuan]

Hi, Can you explain the question in more detail?