search

dmlc / gluon-cv

Gluon CV Toolkit
http://gluon-cv.mxnet.io
Apache License 2.0
5.84k stars 1.22k forks source link

Fine-tuning SlowFast #1745

Closed naynikaw closed 2 years ago

naynikaw commented 2 years ago

I am following this tutorial to fine-tune SlowFast on my own dataset which includes videos at 29 fps, I am using slowfast_4x16_resnet50_custom model with this loader:

train_dataset = VideoClsCustom(root=os.path.expanduser('/content/drive/MyDrive/new'), setting=os.path.expanduser('/content/train.txt'), train=True, new_length=64, slowfast=True, slow_temporal_stride=16, fast_temporal_stride=2, transform=transform_train, video_loader=True, use_decord=True)

but getting the following error while trying to run inference: MXNetError: Check failed: dim_size_is_known(d[i]): Shape dim size must be known, while received -1

Any help would be greatly appreciated. @bryanyzhu

naynikaw commented 2 years ago

fixed the issue by providing the new-length parameter and changing it's value to 64 while running inference.py. closing the issue.