Okay Morganh, Thank you for your big help. I will try to narrow down the resolution and do some experiment again.
Hi m.billson16,
Glad to know you solve the issue.
For mAP(Mean average_precision), it is another topic. Please create a new topic if needed.
But firstly I want to tell you, in the beginning of training, low mAP is expected. It is related to batch-size, epoch, etc.
For detectnet_v2, suggest that you set batch-size= 4 and epoch=120 in your spec file. Monitor the mAP in the end.
Thanks very much for using TLT!
I mean you can decrease the quantity of images/labels. It is not related to resolution.
Okay, Thank you very much again Morganh.
I want to ask if batch-size=4 and epoch=120 is the default value for training?
Okay Morganh, Thank you very much.
Hello Morganh.
I have tried to narrow down my images/labels into 50 images, with num_shards = 2 and val_split=20 and still got the same error.
Traceback (most recent call last):
File "/usr/local/bin/tlt-dataset-convert", line 10, in <module>
sys.exit(main())
File "./detectnet_v2/scripts/dataset_convert.py", line 64, in main
File "./detectnet_v2/dataio/dataset_converter_lib.py", line 74, in convert
File "./detectnet_v2/dataio/dataset_converter_lib.py", line 108, in _write_partitions
File "./detectnet_v2/dataio/dataset_converter_lib.py", line 149, in _write_shard
File "./detectnet_v2/dataio/kitti_converter_lib.py", line 169, in _create_example_proto
File "./detectnet_v2/dataio/kitti_converter_lib.py", line 272, in _add_targets
TypeError: object of type 'int' has no len()
I really don’t understand why these things can happen again. Do you have any idea? I’m very sorry for troubling you
I suggest you add several images/labels based on previous 47 images. If meet error, that means there is something wrong in those image/labels which you newly added.
Okay Morganh, Thank you very much. Thanks for the big help