RCA: Region Conditioned Adaptation for Visual Abductive Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Hao, Ee, Yeo Keat, Fernando, Basura
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916815856205824
author Zhang, Hao
Ee, Yeo Keat
Fernando, Basura
author_facet Zhang, Hao
Ee, Yeo Keat
Fernando, Basura
contents Visual abductive reasoning aims to make likely explanations for visual observations. We propose a simple yet effective Region Conditioned Adaptation, a hybrid parameter-efficient fine-tuning method that equips the frozen CLIP with the ability to infer explanations from local visual cues. We encode "local hints" and "global contexts" into visual prompts of the CLIP model separately at fine and coarse-grained levels. Adapters are used for fine-tuning CLIP models for downstream tasks and we design a new attention adapter, that directly steers the focus of the attention map with trainable query and key projections of a frozen CLIP model. Finally, we train our new model with a modified contrastive loss to regress the visual feature simultaneously toward features of literal description and plausible explanations. The loss enables CLIP to maintain both perception and reasoning abilities. Experiments on the Sherlock visual abductive reasoning benchmark show that the RCA significantly outstands previous SOTAs, ranking the 1st on the leaderboards (e.g., Human Acc: RCA 31.74 $\textit{vs}$ CPT-CLIP 29.58, higher =better). We also validate the RCA is generalizable to local perception benchmarks like RefCOCO. We open-source our project at https://github.com/LUNAProject22/RPA.
format Preprint
id arxiv_https___arxiv_org_abs_2303_10428
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle RCA: Region Conditioned Adaptation for Visual Abductive Reasoning
Zhang, Hao
Ee, Yeo Keat
Fernando, Basura
Computer Vision and Pattern Recognition
Visual abductive reasoning aims to make likely explanations for visual observations. We propose a simple yet effective Region Conditioned Adaptation, a hybrid parameter-efficient fine-tuning method that equips the frozen CLIP with the ability to infer explanations from local visual cues. We encode "local hints" and "global contexts" into visual prompts of the CLIP model separately at fine and coarse-grained levels. Adapters are used for fine-tuning CLIP models for downstream tasks and we design a new attention adapter, that directly steers the focus of the attention map with trainable query and key projections of a frozen CLIP model. Finally, we train our new model with a modified contrastive loss to regress the visual feature simultaneously toward features of literal description and plausible explanations. The loss enables CLIP to maintain both perception and reasoning abilities. Experiments on the Sherlock visual abductive reasoning benchmark show that the RCA significantly outstands previous SOTAs, ranking the 1st on the leaderboards (e.g., Human Acc: RCA 31.74 $\textit{vs}$ CPT-CLIP 29.58, higher =better). We also validate the RCA is generalizable to local perception benchmarks like RefCOCO. We open-source our project at https://github.com/LUNAProject22/RPA.
title RCA: Region Conditioned Adaptation for Visual Abductive Reasoning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2303.10428