Nagabhushan S N

Yes. That is one of their key contributions. They say to 1. use strided convolutions instead of max-pooling layer 2. use cond2dTranspose instead of upsampling 3. use batch normalization layers...

Any luck with getting the depth map?

Check this: [issue 22](https://github.com/alexlee-gk/video_prediction/issues/22) Same problem. Uncomment line 118 in `video_prediction/datasets/kth_dataset.py` and rerun `download_and_preprocess_dataset.sh` I used batch-size=8 and num-samples=816. It has worked for me.

I further tried to get sample predictions. As expected, I got error (see below) `Traceback (most recent call last): File ".../lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow.py", line 58, in from tensorflow.python.pywrap_tensorflow_internal import * File ".../lib/python3.7/site-packages/tensorflow/python/pywrap_tensorflow_internal.py",...

I tried skipping this step. I got the below error `CUDA_VISIBLE_DEVICES=0 python scripts/generate.py --input_dir data/bair --dataset_hparams sequence_length=30 --checkpoint pretrained_models/bair_action_free/ours_savp --mode test --results_dir results_test_samples/bair_action_free loading options from checkpoint pretrained_models/bair_action_free/ours_savp ----------------------------------- Options...

> Use "ffmpeg" command, try "ffmpeg -i vid1.avi vid2.avi" and use vid2.avi to replace vid1.avi. For me, it can solve this problem. May be there are some formats of videos...

I think I found the problem. In my system, in that particular conda environment, there was a python package `ffmpeg` installed. This python package always overrides default ffmpeg installed. So,...

Even i thought so. But this is what I found.

@dnlwbr did you find anything related to your above questions?

Thanks @dnlwbr! This is helpful. I'll update here if I figure out anything more.