DEV Community

0 seconds of 0 secondsVolume 90%
Press shift question mark to access a list of keyboard shortcuts
00:00
00:00
00:00
 
Jimmy Guerrero for Voxel51

Posted on

1

ECCV 2024 Redux: Robust Calibration of Large Vision-Language Adapters

We empirically demonstrate that popular CLIP adaptation approaches, such as Adapters, Prompt Learning, and Test-Time Adaptation, substantially degrade the calibration capabilities of the zero-shot baseline in the presence of distributional drift. We identify the increase in logit ranges as the underlying cause of miscalibration of CLIP adaptation methods, contrasting with previous work on calibrating fully-supervised models. Motivated by these observations, we present a simple and model-agnostic solution to mitigate miscalibration, by scaling the logit range of each sample to its zero-shot prediction logits

ECCV 2024 Paper: Robust Calibration of Large Vision-Language Adapters

About the Speaker: Balamurali Murugesan is currently pursuing his Ph.D. in developing reliable deep learning models. Earlier, he completed his master’s thesis on accelerating MRI reconstruction. He has published 25+ research articles in renowned venues.

Recorded on Nov 19, 2024

DevCycle image

Fast, Flexible Releases with OpenFeature Built-in

Ship faster on the first feature management platform with OpenFeature built-in to all of our open source SDKs.

Start shipping

Top comments (0)

Feature flag article image

Create a feature flag in your IDE in 5 minutes with LaunchDarkly’s MCP server 🏁

How to create, evaluate, and modify flags from within your IDE or AI client using natural language with LaunchDarkly's new MCP server. Follow along with this tutorial for step by step instructions.

Read full post

👋 Kindness is contagious

Explore this insightful write-up embraced by the inclusive DEV Community. Tech enthusiasts of all skill levels can contribute insights and expand our shared knowledge.

Spreading a simple "thank you" uplifts creators—let them know your thoughts in the discussion below!

At DEV, collaborative learning fuels growth and forges stronger connections. If this piece resonated with you, a brief note of thanks goes a long way.

Okay