GenWarp: Single Image to Novel Views with Semantic-Preserving Generative Warping

Authors: Junyoung Seo, Kazumi Fukuda, Takashi Shibuya, Takuya Narihira, Naoki Murata, Shoukang Hu, Chieh-Hsin Lai, Seungryong Kim, Yuki Mitsufuji

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

Reproducibility Variable Result LLM Response
Research Type Experimental Qualitative and quantitative evaluations demonstrate that our model outperforms existing methods in both in-domain and out-of-domain scenarios.
Researcher Affiliation Collaboration Junyoung Seo1,3 Kazumi Fukuda1 Takashi Shibuya1 Takuya Narihira1 Naoki Murata1 Shoukang Hu1 Chieh-Hsin Lai1 Seungryong Kim3 Yuki Mitsufuji1,2 1Sony AI 2Sony Group Corporation 3KAIST AI
Pseudocode No The paper describes the method in text and with diagrams (e.g., Figure 3), but no formal pseudocode or algorithm blocks are provided.
Open Source Code Yes Project page is available at https://Gen Warp-NVS. github.io.
Open Datasets Yes We fine-tune the model on multi-view datasets including indoor scene and outdoor scene, i.e., Real Estate10K [52], Scan Net [8], ACID [25].
Dataset Splits No The paper mentions using Real Estate10K [52], Scan Net [8], and ACID [25] for training, and tests on Real Estate10K and Scan Net. However, it does not explicitly provide specific details for a separate validation split, such as percentages or sample counts.
Hardware Specification Yes We initialize our two networks, semantic preserver and diffusion U-net, with Stable Diffusion v1.5 [35], and fine-tune the networks on 2 H100 80GB with a batch size of 48 for 2-3 days, at resolutions of 512 384 and 512 512.
Software Dependencies Yes We leverage the pretrained Stable Diffusion 1.5 model [35] for both diffusion U-net and semantic preserver network... Stable Diffusion v1.5 Model card: https://huggingface.co/runwayml/stable-diffusion-v1-5
Experiment Setup Yes We initialize our two networks, semantic preserver and diffusion U-net, with Stable Diffusion v1.5 [35], and fine-tune the networks on 2 H100 80GB with a batch size of 48 for 2-3 days, at resolutions of 512 384 and 512 512.