ten1er
commented
4 years ago Hi, I missed similar problem, have you solved it ?
Open duynn912 opened 5 years ago
ten1er
commented
4 years ago Hi, I missed similar problem, have you solved it ?
I successfully trained SNIPER on pascal with this config "sniper_res101_e2e_pascal_voc.yml" and I tried to train a new dataset with this but I have a problem as follow: Traceback (most recent call last): File "main_train.py", line 145, in
epoch_end_callback=epoch_end_callback, arg_params=arg_params, aux_params=aux_params)
File "SNIPER-mxnet/python/mxnet/module/base_module.py", line 520, in fit
self.update_metric(eval_metric, data_batch.label)
File "SNIPER-mxnet/python/mxnet/module/module.py", line 757, in update_metric
self._exec_group.update_metric(eval_metric, labels)
File "SNIPER-mxnet/python/mxnet/module/executor_group.py", line 616, in update_metric
eval_metric.updatedict(labels, preds)
File "SNIPER-mxnet/python/mxnet/metric.py", line 304, in update_dict
metric.update_dict(labels, preds)
File "SNIPER-mxnet/python/mxnet/metric.py", line 132, in update_dict
self.update(label, pred)
File "lib/train_utils/metric.py", line 200, in update
cls = pred[keep_inds, label]
IndexError: index 144 is out of bounds for axis 1 with size 81
Although I have changed the NUM_CLASSES and paths to dataset in "sniper_res101_e2e_pascal_voc.yml", the name of classes in pascal_voc.py. My dataset has 200 classes. So do I need to change anything in other files for training on a new dataset. I also trained with the command to generate proposals but after 2 epoch for proposals, it suffers from the above problem. If someone has a glimpse of this note, please help me out. Thank you so much!