SciCap: Generating Captions for Scientific Figures

Open paper

Benchmark	Model	Rank	Results
image-captioning-on-scicap	CNN+LSTM (Vision only, First sentence)	#1	BLEU-4: 0.0219
image-captioning-on-scicap	CNN+LSTM (Text only, First sentence)	#2	BLEU-4: 0.0213
image-captioning-on-scicap	CNN+LSTM (Text only, Single-Sent Caption)	#3	BLEU-4: 0.0212
image-captioning-on-scicap	CNN+LSTM (Vision only, Single-Sent Caption)	#4	BLEU-4: 0.0207
image-captioning-on-scicap	CNN+LSTM (Vision + Text, First sentence)	#5	BLEU-4: 0.0205
image-captioning-on-scicap	CNN+LSTM (Vision + Text, Single-Sent Caption)	#6	BLEU-4: 0.0202
image-captioning-on-scicap	CNN+LSTM (Vision only, Caption w/ <=100 words)	#7	BLEU-4: 0.0172
image-captioning-on-scicap	CNN+LSTM (Vision + Text, Caption w/ <=100 words)	#8	BLEU-4: 0.0168
image-captioning-on-scicap	CNN+LSTM (Text only, Caption w/ <=100 words)	#9	BLEU-4: 0.0165