Skip to content

Latest commit

 

History

History
158 lines (121 loc) · 4.85 KB

README.md

File metadata and controls

158 lines (121 loc) · 4.85 KB
Raw

singnetlogo

Sequence to Sequence - Video to Text

This service uses S2VT to describe video content with natural language text.

It is part of our third party DNN Model Services.

Getting Started

Requirements

Development

We recommend that you follow the setup from author's repository: https://github.com/vsubhashini/caffe/tree/recurrent/examples/s2vt

You'll need the python Caffe module:

$ nvidia-docker run --name "SNET_VideoCaptioning" -p 7004:7004 -ti bvlc/caffe:gpu bash
  • Inside the container install python-tk and nano:
# apt update
# apt install python-tk nano

Then clone this repository:

# git clone https://github.com/singnet/dnn-model-services.git
# cd dnn-model-services/services/s2vt-video-captioning

Necessary files:

  • You can copy all files from author's repository setup and put them inside service/utils/data.
  • Or, you can use the script service/utils/get_s2vt.sh to download them.
  • You'll need the following files at service/utils/data:
# ls -la service/utils/data
total 882120
-rw-r--r-- 1 user user 553432081 Nov  6 10:59 VGG_ILSVRC_16_layers.caffemodel
-rwxr-xr-x 1 user user      2728 Nov  6 11:05 s2vt.words_to_preds.deploy.prototxt
-rw-r--r-- 1 user user 349436137 Nov  6 11:01 s2vt_vgg_rgb.caffemodel
-rw-r--r-- 1 user user      4582 Nov  6 11:03 vgg_orig_16layer.deploy.prototxt
-rw-r--r-- 1 user user    393710 Nov  6 10:59 yt_coco_mvad_mpiimd_vocabulary.txt

Running the service:

To get the ORGANIZATION_ID and SERVICE_ID you must have already published a service (check this link).

Create the SNET Daemon's config JSON file (snetd.config.json).

{
   "DAEMON_END_POINT": "DAEMON_HOST:DAEMON_PORT",
   "IPFS_END_POINT": "http://ipfs.singularitynet.io:80",
   "BLOCKCHAIN_NETWORK_SELECTED": "BLOCKCHAIN_NETWORK",
   "PASSTHROUGH_ENDPOINT": "http://SERVICE_GRPC_HOST:SERVICE_GRPC_PORT",  
   "ORGANIZATION_ID": "ORGANIZATION_ID",
   "SERVICE_ID": "SERVICE_ID",
   "LOG": {
       "LEVEL": "debug",
       "OUTPUT": {
            "TYPE": "stdout"
           }
   }
}

For example (using the Ropsten testnet):

$ cat snetd.config.json
{
   "DAEMON_END_POINT": "0.0.0.0:7056",
   "IPFS_END_POINT": "http://ipfs.singularitynet.io:80",
   "BLOCKCHAIN_NETWORK_SELECTED": "ropsten",
   "PASSTHROUGH_ENDPOINT": "http://localhost:7003",
   "ORGANIZATION_ID": "snet",
   "SERVICE_ID": "s2vt-video-captioning",
   "LOG": {
       "LEVEL": "debug",
       "OUTPUT": {
           "TYPE": "stdout"
           }
   }
}

Note that we set DAEMON_HOST = 0.0.0.0 because this service will run inside a Docker container.

Install all dependencies:

$ pip2 install -r requirements.txt

Generate the gRPC codes:

$ sh buildproto.sh

Start the service and SNET Daemon:

$ python2 run_service.py

Calling the service:

Inputs:

  • url: An YouTube video URL.
  • start_time_sec: Start time position, in seconds.
  • stop_time_sec: Stop time position, in seconds.
  • The time interval (stop-start) must be <= 20 seconds.

Local (testing purpose):

$ python2 test_service.py 
Endpoint (localhost:7003): 
Method (video_cap): 
Url: http://crcv.ucf.edu/THUMOS14/UCF101/UCF101/v_PlayingGuitar_g05_c01.avi
StartTime(s): 0
StopTime (s): 0
{'Caption': '1\n00:00:00,00 --> 00:00:10,00\nA man is playing guitar.'}

Through SingularityNET (follow this link to learn how to publish a service and open a payment channel to be able to call it):

Assuming that you have an open channel to this service:

$ snet client call snet s2vt-video-captioning default_group video_cap '{"url": "http://crcv.ucf.edu/THUMOS14/UCF101/UCF101/v_PlayingGuitar_g05_c01.avi", "start_time_sec": "0", "stop_time_sec": "0"}'
...
Read call params from cmdline...

Calling service...

    response:
        value: '{''Caption'': ''1\n00:00:00,00 --> 00:00:10,00\nA man is playing guitar.''}'

Contributing and Reporting Issues

Please read our guidelines before submitting an issue. If your issue is a bug, please use the bug template pre-populated here. For feature requests and queries you can use this template.

Authors