search

ejcgt / attention-target-detection

[CVPR2020] "Detecting Attended Visual Targets in Video"
MIT License
173 stars 48 forks source link

No such file or directory: 'model_demo.pt' #5

Open zbeedatm opened 4 years ago

zbeedatm commented 4 years ago

I downloaded the code to my machine and tried to run it from PyCharm (with Conda python after installing torch -on Windows without GPU-) but when running the demo.py I'm getting:

No such file or directory: 'model_demo.pt'

where can I get that file from? Is it possible to run it on windows this way? and without GPU... will it work?

zbeedatm commented 4 years ago

I downloaded the .pt files manually, took their urls from download_models.sh:

wget https://www.dropbox.com/s/s9y65ajzjz4thve/initial_weights_for_spatial_training.pt wget https://www.dropbox.com/s/ye3lyyyzd73afa7/initial_weights_for_temporal_training.pt wget https://www.dropbox.com/s/vt8hua06r1yoi2i/model_demo.pt wget https://www.dropbox.com/s/nloym5bmvv1v7wr/model_gazefollow.pt wget https://www.dropbox.com/s/ywd16kcv06vn93x/model_videoatttarget.pt

and downlded the data sets of gazefollow and videoattarget (from links in wiki)

1) demo.py ran successfuly 2) eval_on_gaze_follow is giving this error that I couldn't figure how to fix it yet!

usage: _jb_nosetest_runner.py [-h] [--device DEVICE] [--model_weights MODEL_WEIGHTS] [--batch_size BATCH_SIZE] _jb_nosetest_runner.py: error: unrecognized arguments

3) eval_on_videoatttarget is giving this error:

RuntimeError: CUDA error: out of memory

ejcgt commented 4 years ago

Hi @zbeedatm sorry for my late reply. Were you able to resolve this issue? Are you trying to run the code without GPU? I haven't tested the code with CPU only mode so it may not work as-is if you don't have GPU.