About -cocoDir and -cocoInfo

[okisy]

When I tried to run evaluate.py with “-evalMode dialog” like this
python evaluate.py -useGPU \ -startFrom checkpoints/abot_rl_ep20.vd \ -qstartFrom checkpoints/qbot_rl_ep20.vd \ -evalMode dialog \ -beamSize 5
, I had an error as follows

[Error] Need coco directory and info as input to -cocoDir and -cocoInfo arguments for locating coco image files.
Exiting dialogDump without saving files.

What should I do with this? Is it appropriate for me to make a path to the directory which has MSCOCO dataset?

In addition what are these two(-cocoDir, -cocoInfo) used for?

[nirbhayjm]

Will add information about how to use the coco information in visualization soon.

Briefly, the -cocoDir argument is supposed to be the directory containing all coco images and coco.json should have a dictionary of paths within -cocoDir, which should look something like this:

{"images":
   [
      {
         "id": 152328,
         "split": "val",
         "file_path": "train2014/COCO_train2014_000000152328.jpg"
      },
      {
         "id": 82551,
         "split": "val",
         "file_path": "val2014/COCO_val2014_000000082551.jpg"
      },
      ...
   ]
}

I will look into adding the json file itself to the repository if the mapping to coco images is not obvious in the current data.

[niwtr]

Any update on this one please?

[JackFurby]

I think I have made a script to generate the coco.json file. I have made this into a gist for easy access https://gist.github.com/JackFurby/249c5f584c2df4dda80361af5b05a368. To the best of my knowledge this is correct and the generated JSON file seems to be working when training the models.