Bringing together Black women and allies working, studying, and/or advocating for equity and Public Health. Centering, celebrating, and contributing to Black people, culture, organizations, and institutions. Building capacity for health and equity for Black women, Public Health, and the communities we reach.

BLIPH is now an active network of over 16,000 students, community partners, and professionals in Public Health (PH) and a resource to the community at large. Our pillars for success include scholarship, service, self-care, support, safe space, social innovation, and sustainable pathways to achieve equity in health and the public health profession.

About BLiPH; Leadership; Programs & Partnerships. Black Men In Public Health; Black Girls Spring for R&R; HBCU Response to COVID-19; HealTHY HBCU; High School Community Health Worker Program; Reflections – HIV Program; Events. Gallery; Contact; Donation; 682-391-9889. Login. Login. Username or E-Mail. Password. Forget Password? Do not have an ...

BLiPH is a network and resource for advancing public health careers and outcomes for Black Women and the communities we serve. BLIPH is a registered 501(C)(3) non-profit organization.

BLIPH is now an active network of over 16,000 students, community partners, and professionals in Public Health (PH) and a resource to the community at large. Our pillars for success include …

BLIP-2 由预训练的Image Encoder，预训练的Large Language Model，和一个可学习的 Q-Former 组成。 Image Encoder：从输入图片中提取视觉特征，尝试了两种网络结构，CLIP 训练的 ViT-L/14和EVA-CLIP训练的 ViT-g/14（去掉了最后一层）。 Large Language Model：大语言模型进行文本生成，尝试了接入decoder-based LLM 和 encoder-decoder-based LLM两种结构。 Q-Former：弥补视觉和语言两种模态的modality gap，可以理解为固定图像编码器和固定LLM之 …

4,528 Followers, 1,584 Following, 380 Posts - Black Ladies in Public Health (@Bliph16) on Instagram: "We aim to build and promote positive experiences and narratives of Black Ladies in Public Health! #bliph BLIPH ON CLUBHOUSE: www.bepublichealth.com"

BLIP系列是多模态任务比较有代表性的一份工作，本文将对BLIP系列的3篇paper进行详细解读，便于多模态的初学者入门，也以便自己日后回顾。 概括来看： BLIP的核心创新点在于boostrapping caption的方案的设计。 该方案用于“提纯”带噪声web datasets，从而进一步提升多模态模型的效果。 BLIP-2的核心创新点有二，其一是设计了一个轻量架构 QFormer （querying transformer）来建立图像-文本的桥梁，其二是设计了一种二阶段预训练范式实现高效训练，实现将目前的视 …

2024年3月19日 · BLiPH's mission is to connect, celebrate, and contribute to the health and lives of Black women, promoting sustainable pathways for equity in wealth and health through donations, sponsorships, and community engagement.

BLIP-2 在多模态大模型领域具有深远的意义，它提出的freeze ViT + LLM，仅训练少量的connector的模式被后续的大量的工作所应用（例如LLaVA、MiniGPT等）。 它也获得了应有的影响力，截止撰文时已经有超过1.1K个citations和超过8K的GitHub stars。 今天我们一起回顾下这篇经典之作。 Li J, Li D, Savarese S, et al. Blip-2: Bootstrapping language-image pre-training with frozen image encoders and large language models.