AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Self-Adapting Large Visual-Language Models to Edge Devices across Visual Modalities

Kaiwen Cai, ZheKai Duan, Gaowen Liu, Charles Fleming, Chris Xiaoxuan Lu

ECCV 2024 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add contrastive learning to quantization-aware training (stage-2) instead of plain QAT (ablation)EdgeVL DAT-T · ScanNetavg top-1 accuracy (%): 40.3 -> 50.0Improved
apply post-training quantization to the stage-1 student (ablation)EdgeVL ViT-S · ScanNetavg top-1 accuracy (%): 39.3 -> 25.3Hurt
train a dual-modality student instead of single-modality CMKD (non-RGB) (ablation)EdgeVL stage-1 · ScanNetavg top-1 accuracy (%): 24.6 -> 39.6Improved
use RGB images as extra training signal for the non-RGB branch (ablation)EdgeVL stage-1 · ScanNetnon-RGB accuracy (%): 37.8 -> 38.6Improved
set automatic curation threshold tau_c to 0.25 instead of 0.10 (ablation)EdgeVL Swin-T · EuroSATavg top-1 accuracy (%): 51.4 -> 64.2Improved
raise tau_c from 0.25 to 0.50 (ablation)EdgeVL Swin-T · EuroSATavg top-1 accuracy (%): 64.2 -> 62.0Hurt
use semi-hard instead of hard triplet sampling (ablation)EdgeVL DAT-T · EuroSATavg top-1 accuracy (%): 62.7 -> 63.3Improved
use two-stage training instead of one-stage with PTQ (ablation)EdgeVL DAT-T · ScanNetavg top-1 accuracy (%): 30.0 -> 49.9Improved
use EdgeVL ViT-S Int8 instead of CLIP-B on Jetson AGXedge inference · Jetson AGX Orinlatency (ms): 9.5 -> 4.6Improved
adapt on ScanNet and test on NYU2 RGB images vs pretrained CLIP-GEdgeVL DAT-T · NYU2RGB accuracy (%): 69.7 -> 54.3Hurt
adapt on ScanNet and test depth images on NYU2 vs pretrained CLIP-GEdgeVL DAT-T · NYU2non-RGB accuracy (%): 25.7 -> 51.1Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy