aimagelab / show-control-and-tell

Show, Control and Tell: A Framework for Generating Controllable and Grounded Captions. CVPR 2019
https://arxiv.org/abs/1811.10652
BSD 3-Clause "New" or "Revised" License
282 stars 62 forks source link

Got some trouble with install speaksee #13

Closed AlanHuang1998 closed 4 years ago

AlanHuang1998 commented 5 years ago

Thanks for sharing, I got ERROR when I install the requirement torch, the situation is the same as seen in the photo. image

Is there any way to solve it?

marcellacornia commented 5 years ago

Hi @AlanHuang1998, did you solve the problem?

AlanHuang1998 commented 5 years ago

@marcellacornia Solved it, but got another problem, when I use "python test_region_set.py --dataset coco --exp_name ours --sample_rl_nw" to run, looks like take too long time to download something, so the server shutdown the connection, photo below. 1572336749494

I'm rookie in image caption, want to ask how to train the model to get sentences. Just like you demonstrated.

marcellacornia commented 5 years ago

The tgz file is needed to calculate the spice metric. If the download is too slow, you can comment the two lines of code that calculate the spice score.

val_spice, _ = Spice().compute_score(gts_t, gen_t)
print('SPICE', val_spice)
AlanHuang1998 commented 5 years ago

Thanks for your help! I follow your instructions and get each score. Want to ask how to use the model to get a full sentences, like the picture below. Thank you! 1573650778679

tejasri19 commented 4 years ago

@marcellacornia @AlanHuang1998 When I run python test_region_set.py --dataset coco --exp_name ours --sample_rl_nw, i get import errors of speaksee.data. However the installation shows successful. Screenshot_20200428_153843