Abstract: Text-to-image person re-identification aims to utilize textual descriptions to retrieve specific person images from large image databases. The core challenge of this task lies in the ...
If you just want to use MIR as the pre-training indicator of your own model, no additional environment is required. python mir.py --model_path PATH/TO/MODEL --base_llm PATH/TO/LLM --text_data_path ...
Semantic Entity Alignment and Non-Corresponding Reasoning for Text-to-Image Person Re-identification
Abstract: With the rapid development of intelligent surveillance technology, the massive amount of multimodal data (e.g., videos, images, and text) has imposed higher demands on efficient information ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results