BLIP
BLIP (Bootstrapping Language-Image Pre-training) is a vision-language pretraining framework developed by Salesforce Research that unifies understanding and generation tasks, and introduces a 'CapFilt' bootstrapping method that uses a captioner and filter to clean noisy web-scraped image-text data during training.
3 resources across 1 library
Glossary Terms(3)
BLIP
BLIP (Bootstrapping Language-Image Pre-training) is a vision-language pretraining framework developed by Salesforce Research that unifies understanding and gen…
BLIP-2
BLIP-2 is a vision-language pretraining framework from Salesforce Research that connects frozen, pretrained image encoders and frozen large language models usi…
InstructBLIP
InstructBLIP is a vision-language model from Salesforce Research that extends BLIP-2's frozen-encoder, Q-Former architecture with instruction-tuning, training…