Skip to content
/ updown-baseline Public template

Baseline model for nocaps benchmark, ICCV 2019 paper "nocaps: novel object captioning at scale".

License

Notifications You must be signed in to change notification settings

nocaps-org/updown-baseline

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

59 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

UpDown Captioner Baseline for nocaps

Baseline model for nocaps benchmark, a re-implementation based on the UpDown image captioning model trained on the COCO dataset (only), and with added support of decoding using Constrained Beam Search.

predictions generated by updown model

Citation

If you find this code useful, please consider citing our paper, the paper which proposed original model, and EvalAI — the platform which hosts our evaluation server. All bibtex available in CITATION.md.

Usage Instructions

  1. How to setup this codebase?
  2. How to train your captioner?
  3. How to evaluate or run inference?

Extensive documentation available at nocaps.org/updown-baseline. Use it as an API reference to navigate through and build on top of our code.

Results

Pre-trained checkpoints with the provided configs in (configs directory) are available to download:

UpDown Captioner (no CBS):

Note: While CBS is inference-only technique, it cannot be used on this checkpoint. CBS requires models to have 300-dimensional froze GloVe embeddings, this checkpoint has 1000- dimensional word embeddings which are learned during training.

in-domain near-domain out-of-domain overall
CIDErSPICE CIDErSPICE CIDErSPICE BLEU1BLEU4METEORROUGECIDErSPICE
78.111.6 57.710.3 31.38.3 73.718.322.750.455.310.1

UpDown Captioner + Constrained Beam Search:

Note: Since CBS is inference-only technique, this particular checkpoint can be used without CBS decoding. It yields similar results to the UpDown Captioner trained using learned word embeddings during training.

With CBS Decoding:

in-domain near-domain out-of-domain overall
CIDErSPICE CIDErSPICE CIDErSPICE BLEU1BLEU4METEORROUGECIDErSPICE
78.612.1 73.511.5 68.89.8 75.817.522.751.173.311.3

Without CBS Decoding:

in-domain near-domain out-of-domain overall
CIDErSPICE CIDErSPICE CIDErSPICE BLEU1BLEU4METEORROUGECIDErSPICE
75.711.7 58.010.3 32.98.2 73.118.022.750.255.410.1