Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

TSP-Transformer: Task-Specific Prompts Boosted Transformer for Holistic Scene Understanding

About

Holistic scene understanding includes semantic segmentation, surface normal estimation, object boundary detection, depth estimation, etc. The key aspect of this problem is to learn representation effectively, as each subtask builds upon not only correlated but also distinct attributes. Inspired by visual-prompt tuning, we propose a Task-Specific Prompts Transformer, dubbed TSP-Transformer, for holistic scene understanding. It features a vanilla transformer in the early stage and tasks-specific prompts transformer encoder in the lateral stage, where tasks-specific prompts are augmented. By doing so, the transformer layer learns the generic information from the shared parts and is endowed with task-specific capacity. First, the tasks-specific prompts serve as induced priors for each task effectively. Moreover, the task-specific prompts can be seen as switches to favor task-specific representation learning for different tasks. Extensive experiments on NYUD-v2 and PASCAL-Context show that our method achieves state-of-the-art performance, validating the effectiveness of our method for holistic scene understanding. We also provide our code in the following link https://github.com/tb2-sy/TSP-Transformer.

Shuo Wang, Jing Li, Zibo Zhao, Dongze Lian, Binbin Huang, Xiaomei Wang, Zhengxin Li, Shenghua Gao• 2023

Related benchmarks

TaskDatasetResultRank
Surface Normal EstimationNYU v2 (test)--
224
Depth EstimationNYU Depth V2
RMSE0.4961
209
Semantic segmentationNYUD v2
mIoU55.39
125
Saliency DetectionPascal Context (test)
maxF84.86
57
Surface Normal EstimationPascal Context (test)
mErr13.69
50
Boundary DetectionPascal Context (test)
ODSF74.8
34
Boundary DetectionNYUD v2
ODS F-measure77.5
30
Human Part ParsingPascal Context (test)
mIoU70.64
20
Showing 8 of 8 rows

Other info

Follow for update