search

nihui / dain-ncnn-vulkan

DAIN, Depth-Aware Video Frame Interpolation implemented with ncnn library
MIT License
515 stars 45 forks source link

output frames failed #27

Closed meensukt closed 3 years ago

meensukt commented 3 years ago

I just followed the Tutorial and failed to interpolate frames from this command:

./dain-ncnn-vulkan -i input_frames -o output_frames

The input_frames folder is full of png files named frame_000xx.png The output_frames folder is originally empty but after the command, I get the input frames alternating with black images.

The error log in the terminal says:

find_blob_index_by_name input failed find_blob_index_by_name depth failed find_blob_index_by_name input failed find_blob_index_by_name depth failed find_blob_index_by_name input0 failed find_blob_index_by_name input1 failed find_blob_index_by_name flow failed find_blob_index_by_name input0 failed find_blob_index_by_name input1 failed find_blob_index_by_name flow failed find_blob_index_by_name input failed find_blob_index_by_name ctx failed find_blob_index_by_name input failed find_blob_index_by_name ctx failed find_blob_index_by_name input0 failed find_blob_index_by_name input1 failed find_blob_index_by_name depth0 failed find_blob_index_by_name depth1 failed find_blob_index_by_name flow0 failed find_blob_index_by_name flow1 failed find_blob_index_by_name flow0_w failed find_blob_index_by_name flow1_w failed find_blob_index_by_name ctx0 failed find_blob_index_by_name ctx1 failed find_blob_index_by_name output_rectified failed

zocker-160 commented 3 years ago

I had the same issue, but I solved it by moving cain.bin and cain.param out of the cain folder and placing them right next to the executable This should fix the problem.

EDIT: I confused the CAIN and DAIN repos, you need to copy the contents of the best folder next to the executable

dexeonify commented 3 years ago

I had the same issue, but I solved it by moving cain.bin and cain.param out of the cain folder and placing them right next to the executable This should fix the problem.

This is DAIN, not CAIN, unless I'm missing something here? I tried to grab them from nihui/cain-ncnn-vulkan and neither does that fix the problem. I don't expect it to fix the issue anyways since it's two different repositories.

zocker-160 commented 3 years ago

@dexeonify oh sorry I confused the two, but the fix still works, since both repos use the same basic code The executable does not look inside the folder for the pretrained model files.

You only need to copy the contents of the best folder next to the executable

dexeonify commented 3 years ago

Isn't that what pull request #20 is supposed to fix?

Edit: If you want to use the latest action build then you would have to use the workaround, since the pr hasn't been merged yet.

zocker-160 commented 3 years ago

Isn't that what pull request #20 is supposed to fix?

yes exactly

nihui commented 3 years ago

fixed in https://github.com/nihui/dain-ncnn-vulkan/commit/f9f849034bf85bb9828bccc68688ca612da8978b

QuantumAnomaly commented 1 year ago

I just downloaded the most recent macOS release, but I'm having this problem. Moving the contents of the "best" folder next to the executable did not fix it. If this was patched in 2020, why is the problem still persisting?