Jul 24, 2024 · We propose an occluded pedestrian recognition model based on diversified Transformer with adaptive constraint (DTAC).
Mar 3, 2023 · We propose a Feature Completion Transformer (FCFormer) to implicitly complement the semantic information of occluded parts in the feature space.
Missing: Adaptive Constraints.
Dec 1, 2023 · We introduce a novel Re-ID model that learns discriminative features on both local and global scales for occluded pedestrian identification.
Missing: Diversified Constraints.
Diversified Transformer For Occluded Person Re Identification Based on Adaptive Constraints. Conference Paper. Jul 2024. Tingting Yuan · Sibao Chen · View.
Feb 16, 2024 · As a result, it generates occlusion images that closely resemble real-world occlusions, greatly enhancing the subsequent contrastive learning ...
Nov 8, 2024 · To tackle the high resource consumption in occluded person re-identification, sparse attention mechanisms based on Vision Transformers ...
Missing: Diversified | Show results with:Diversified
Occluded person re-identification (Re-ID) aims to address the potential occlusion problem when matching occluded or holistic pedestrians from different camera ...
Missing: Diversified | Show results with:Diversified
Oct 22, 2024 · We propose a novel approach named Instant Pose Extraction based on Mask Transformer (MTIPE), tailored specifically for occluded person Re-ID.
To address these issues, we propose a novel end-to-end Part-Aware Transformer (PAT) for occluded person Re-ID through diverse part discovery via a transformer ...
Currently, occluded person Re-ID methods can be broadly categorized into three types: CNN-based methods, external cues-based methods, and Transformer-based ...