Conditional Controllable Image Fusion

Authors: Bing Cao, Xingxin Xu, Pengfei Zhu, Qilong Wang, Qinghua Hu

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments validate our effectiveness in general fusion tasks across diverse scenarios against the competing methods without additional training.
Researcher Affiliation Academia Bing Cao1,2 Xingxin Xu3 Pengfei Zhu1 Qilong Wang1 Qinghua Hu1 1College of Intelligence and Computing, Tianjin University, Tianjin, China 2State Key Laboratory of Integrated Services Networks, Xidian University, Xi an, China 3School of New Media and Communication, Tianjin University, Tianjin, China
Pseudocode Yes Algorithm 1 CCF Input: : i,v Output: : f
Open Source Code Yes The code is publicly available. https://github.com/jehovahxu/CCF
Open Datasets Yes For multi-modal image fusion task, we conducted experiments on the LLVIP [41] dataset and referred to the test set outlined in Zhu et al. [34]. For MEF and MFF, our testing procedure followed the test setting in MFFW dataset [42] and MEFB dataset [43], respectively. Additionally, we test our method on the TNO dataset and Harvard medical dataset to assess our method s performance within the multi-modal fusion domain, detailed in App. B and H.
Dataset Splits Yes For multi-modal image fusion task, we conducted experiments on the LLVIP [41] dataset and referred to the test set outlined in Zhu et al. [34]. For MEF and MFF, our testing procedure followed the test setting in MFFW dataset [42] and MEFB dataset [43], respectively.
Hardware Specification Yes The experiments are conducted on Huawei Atlas 800 Training Server with CANN and NVIDIA RTX 3090 GPU.
Software Dependencies No Our method utilizes a pre-trained diffusion model as our foundational model [44]. The experiments are conducted on Huawei Atlas 800 Training Server with CANN and NVIDIA RTX 3090 GPU. The detection model employed is YOLOv5 [48], trained on the LLVIP dataset.
Experiment Setup Yes Our method utilizes a pre-trained diffusion model as our foundational model [44]. This model was directly applied without any subsequent fine-tuning for specific task requirements during our experiments. The experiments are conducted on Huawei Atlas 800 Training Server with CANN and NVIDIA RTX 3090 GPU. Experimental settings are shown in App. A.