Pointnet_Pointnet2_pytorch
Pointnet_Pointnet2_pytorch copied to clipboard
a correction
I think the 3 lines in train_classification.py:
class_acc[cat, 0] += classacc.item() / float(points[target == cat].size()[0])
class_acc[cat, 1] += 1
shoud be replaced by:
class_acc[cat, 0] += classacc.item()
class_acc[cat, 1] += float(points[target == cat].size()[0])
This will match the code below:
class_acc[:, 2] = class_acc[:, 0] / class_acc[:, 1]
class_acc = np.mean(class_acc[:, 2])