img_cat_cond referenced before assignment error

[jes-bro]

        else:
            ## simply repeat the cond_frame to match the seq_len of z
            # img_cat_cond = z_cat[:,:,cond_frame_index,:,:]
            img_cat_cond = img_cat_cond.unsqueeze(2) # not defined before this point
            img_cat_cond = repeat(img_cat_cond, 'b c t h w -> b c (repeat t) h w', repeat=z.shape[2])

Here is the error message:

        Mask2IV/main/../lvdm/models/ddpm3d_maskgen.py", line 1158, in get_batch_input
            img_cat_cond = img_cat_cond.unsqueeze(2)
        UnboundLocalError: local variable 'img_cat_cond' referenced before assignment

Going to try to figure out what img_cat_cond is supposed to be but would be helpful if this got fixed. Thanks!

[jes-bro]

Hello, in the tensorboard, I wanted to clarify what -samples versus -reconst is. Thanks!

[Reagan1311]

Hi, it seems I forgot to remove the comments # img_cat_cond = z_cat[:,:,cond_frame_index,:,:]. Simply removing the # in this line shold be fine.

Samples are the generated videos, and reconst are the GT videos

[jes-bro]

Thank you! I also have one more question. In run_hoi4d_2stage, what should prompt_dir point to? The data_csv directory?

[Reagan1311]

This is for inference not for benchmark evaluation, so you'll need to prepare the input image and corresponding object mask in the prompt_dir folder to do inference. You can check the inference_hoi4d_2stage.py for more details.