Closed ypzuo closed 4 years ago
I changed the parameter to 'pcl_gt_scaled', it worked. But another problem appeared, in the 'feed_dict={pcl_gt:batch_gt, img_inp:batch_ip}' showed the problem is 'ValueError: cannot feed value of shape(0,) for tensor u'pcl_gt:0',which has hsape '(24,1024,3) ' What should I do ?
I changed the parameter to 'pcl_gt_scaled', it worked. But another problem appeared, in the 'feed_dict={pcl_gt:batch_gt, img_inp:batch_ip}' showed the problem is 'ValueError: cannot feed value of shape(0,) for tensor u'pcl_gt:0',which has hsape '(24,1024,3) ' What should I do ?
Hello, did u find the solution ?
I changed the parameter to 'pcl_gt_scaled', it worked. But another problem appeared, in the 'feed_dict={pcl_gt:batch_gt, img_inp:batch_ip}' showed the problem is 'ValueError: cannot feed value of shape(0,) for tensor u'pcl_gt:0',which has hsape '(24,1024,3) ' What should I do ?
I changed the parameter to 'pcl_gt_scaled', it worked. But another problem appeared, in the 'feed_dict={pcl_gt:batch_gt, img_inp:batch_ip}' showed the problem is 'ValueError: cannot feed value of shape(0,) for tensor u'pcl_gt:0',which has hsape '(24,1024,3) ' What should I do ?
Hello, did u find the solution ?
Sorry for bothering, Do you guys find any solution?
I solve the above two problems' by changing lines: 282, 283, 285, 286, 290 in metrics.py
:
273: batches = len(train_pair_indices)
274: calculate_metrics(train_models, batches, pcl_gt_1K_scaled, reconstr_img_scaled, train_pair_indices)
276: batches = len(val_pair_indices)
277: calculate_metrics(val_models, batches, pcl_gt_1K_scaled, reconstr_img_scaled, val_pair_indices)
281: batches = len(models)
into
import math
273: batches = int(math.ceil(len(train_pair_indices) / BATCH_SIZE))
274: calculate_metrics(train_models, batches, pcl_gt_scaled, reconstr_img_scaled, train_pair_indices)
276: batches = int(math.ceil(len(val_pair_indices) / BATCH_SIZE))
277: calculate_metrics(val_models, batches, pcl_gt_scaled, reconstr_img_scaled, val_pair_indices)
281: batches = int(math.ceil(len(models) / BATCH_SIZE))
I think the author forgot to deal with batch size problems. It seems more reasonable to divide the length of the dataset with batch size since it later iterates through the batches in function: calculate_metrics.
As for changing pcl_gt_1K_scaled to pcl_gt_scaled, I'm not sure what the author's intention is here and how it would affect the performance. But at least the Evaluation part in README.md can run smoothly now.
If there is any misunderstanding of the code, please let me know. Thx!
Hi, I want use the evaluation to see the cd, emd etc. But it report the issue. I see some file but didn't find the detail. I miss something here?