A learnable-gate pruning method for image captioning decoders reaches 97.5% sparsity (40x compression) with roughly 2% BLEU-4 and CIDEr loss after end-to-end fine-tuning on MS-COCO.
Persistent RNNs: Stashing recurrent weights on-chip,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Image Captioning with Sparse Recurrent Neural Network
A learnable-gate pruning method for image captioning decoders reaches 97.5% sparsity (40x compression) with roughly 2% BLEU-4 and CIDEr loss after end-to-end fine-tuning on MS-COCO.