Abstract
Abstract Capitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine
Read MoreAbstract Capitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine
Read MoreST-Adapter: Parameter-Efficient Image-to-Video Transfer Learning Junting Pan*, Ziyi Lin*, Xiatian Zhu, Jing Shao, Hongsheng Li
Read MoreContribute to linziyi96/st-adapter development by creating an account on GitHub.
Read MoreSummary: This paper proposes a new Spatio-Temporal Adapter (ST-Adapter) for parameter-efficient fine-tuning on video tasks. With a much smaller trainable parameter, ST-Adapter
Read MoreThe experiments show that the ST-Adapter performs as well as or even better than state-of-the-art video models while requiring less training time and computational resources.
Read MoreThis work proposes a new Spatio-Temporal Adapter (ST-Adapter) for parameter-efficient fine-tuning per video task, with a built-in spatio-temporal reasoning capability in a compact design, that enables a
Read More4. Ablation Study on Efficiency The same ViT-B/16 with CLIP pre-training is used for all experiments. Models & source code: https://github /linziyi96/st-adapter
Read MoreWe release the data list we used for Kinetics-400 (k400, train list link, val list link) and Something-something-v2 (ssv2, train list link, val list link), which reflect the class mapping of the released
Read MoreWith a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video
Read MoreIn this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this
Read MoreWith a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video content at a small
Read MoreWith a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video content at a small
Read MoreAbstract Capitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine
Read Morewe also show the performance impact of using fewer ST-Adapters. As shown in Table 5b, while more ST-Adapters tend to do better, ST-Adapters at deeper layers boost
Read MoreIn this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this problem, we propose a new Spatio-Temporal
Read MoreIn this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this problem, we propose a new...
Read MoreCapitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full
Read MoreBoth the Standard 4 X and Standard Actuated Kit provide the same reliable, high-speed internet you can expect from Starlink. Both of these Standard hardware
Read MoreCapitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full
Read MoreCapitalizing on large pre-trained models for various downstream tasks of interest have recently emerged with promising performance. Due to the ever-growing model size, the standard full fine-tuning based
Read MoreIn this work, we investigate such a novel cross-modality transfer learning setting, namely parameter-efficient image-to-video transfer learning. To solve this problem, we propose a new
Read MoreWith a built-in spatio-temporal reasoning capability in a compact design, ST-Adapter enables a pre-trained image model without temporal knowledge to reason about dynamic video
Read More+27 10 247 8396
Unit 7, Summit Place, 21 Summit Rd, Midrand, Johannesburg, 1685, South Africa