Errors running the image

[ir0nt0ad]

I'm getting errors like:

Number of GPUs present 1
Device 0:  "GeForce RTX 2070 with Max-Q Design"  7982Mb, sm_75, Driver/Runtime ver.10.20/8.0
TVL1 parameters
- epsilon: 0.01
- lambda: 0.15
- tau: 0.25
- theta: 0.3
- iterations: 300
- nscales: 5
- warps: 5
- use_initial_flow: 0
Processing folder /input
/bin/compute_flow_wrapper.sh: line 27:     9 Killed                  "$COMPUTE_FLOW" "$IN" "$FRAME_PATTERN" "$OF_PATTERN" $@

The u and v folders are being created, but there's nothing inside.
Btw, I changed --runtime=nvidia to --gpus all in the script; it also says just docker instead of docker run. My filenames look like frame000001.jpg, so I'm using frame%06d.jpg.

Ubuntu 18.04, Docker 19.03.8

[willprice]

Can you give me the exact command you're using to invoke the application. Ideally as a docker command defining the mounts and a description of what those folder hierarchies look like, then I'll be able to help debug

[ir0nt0ad]

Hi Will,

Apologies for the delay, just found some time to get back to the issue. My current setup as simple as it could be, a folder with frames called jpegs and an empty directory called flow. The frames are named like frame000001.jpg. I run the following command:

docker run \
    --gpus all \
    --rm \
    --mount "type=bind,source=$PWD/jpegs,target=/input" \
    --mount "type=bind,source=$PWD/flow,target=/output" \
    --mount "type=bind,source=$HOME/.nv,target=/cache/nv" \
    willprice/furnari-flow \
    frame%06d.jpg

It appears that I set the input file format correctly, otherwise I'm getting an error along the lines of [image2 @ 0xfd9620] Could find no file with path '/input/frame%07d.jpg' and index in the range 0-4.

[willprice]

Hi Roman,
Hmmm, that's weird. I've taken the command you wrote and I was able to successfully run it on a system based on Ubuntu 18.04 and docker 19.03.8. I dumped frames from big buck bunny from https://test-videos.co.uk/

$ wget https://test-videos.co.uk/vids/bigbuckbunny/mp4/h264/1080/Big_Buck_Bunny_1080_10s_30MB.mp4
$ ffmpeg -i Big_Buck_Bunny_1080_10s_30MB.mp4 jpegs/frame%06d.jpg
$ docker run \
    --gpus all \
    --rm \
    --mount "type=bind,source=$PWD/jpegs,target=/input" \
    --mount "type=bind,source=$PWD/flow,target=/output" \
    --mount "type=bind,source=$HOME/.nv,target=/cache/nv" \
    willprice/furnari-flow \
    frame%06d.jpg
Number of GPUs present 1
Device 0:  "GeForce GTX 1060 6GB"  6078Mb, sm_61, Driver/Runtime ver.10.20/8.0
TVL1 parameters
- epsilon: 0.01
- lambda: 0.15
- tau: 0.25
- theta: 0.3
- iterations: 300
- nscales: 5
- warps: 5
- use_initial_flow: 0
Processing folder /input
..........................................................................................................................................................................................................................................................................................................

It's possibly an issue with having a 20-series GPU, but since I don't have any of those I'm unable to test. Maybe try with big buck bunny just to check its not a problem with your media.

[ir0nt0ad]

Hi, got the same error as before unfortunately :( must be my GPU indeed.