The First Workshop on Efficient and On-Device Generation (EDGE) at CVPR 2024 will focus on the latest advancements of generative AI in the computer vision domain, with an emphasis on efficiencies across multiple aspects. We encourage techniques that enable generative models to be trained more efficiently and/or run on resource-constrained devices, such as mobile phones and edge devices. Through these efforts, we envision a future where these permeating generative AI capabilities become significantly more accessible with virtuous scalability and plateauing carbon footprint.

The topics involved in the workshop include but are not limited to:

• Training efficiency aspect:
• Methods for reducing the memory footprint of generative models.
• Using knowledge distillation to support model compression.
• Design of efficient computational modules and algorithms for generative models
• Hardware-aware training acceleration and energy efficient computing.
• Online training under edge constraints.
• Parameter efficient tuning.
• Inference efficiency aspect:
• Techniques for reducing the computational complexity of generative models, such as model compression, quantization, sparsification, and distillation.
• Advanced sampling techniques for fast inference.
• Approaches for efficient deployment and distribution generative models across multiple devices or edge devices.
• Hardware-aware inference acceleration and energy efficient computing.
• Data efficiency aspect:
• Few-shot diffusion model.
• Generative model finetuning.
• Data efficiency for generative models and/or foundation models.
• Evaluation efficiency aspect:
• Efficient evaluation of generative models.
• Methods for benchmarking the efficiency of generative models.
• Challenges in evaluating the efficiency of generative models in diverse applications.
• Application-specific efficiency aspect:
• Applications of efficient generative AI in computer vision, such as image generation, style transfer, and super-resolution.
• Efficient personalization and post-generation editing.
• Real-time and efficient neural rendering.
• Efficient novel view synthesis.
• Interdisciplinary approaches that combine efficient generative AI with other fields, such as natural language processing, robotics, and human-computer interaction.
• Efficient generative models for video synthesis.
• Efficient interplay between image generation, video generation, and language generation such as efficient text2img, img2text, text2video, video2text, img2video, video2img, etc.
• Other related topics:
• Challenges and opportunities.
• Exploratory direction.
• New evaluation metrics.
• Privacy considerations.

Format: Submissions must use the CVPR 2024 Author Kit for Latex/Word Zip file and follow the CVPR 2024 author instructions and submission policies. Submissions need to be anonymized. The workshop considers two submission tracks:

• Long Paper: submissions are limited to 8 pages excluding references;
• Extended Abstract: submissions are limited to 4 pages excluding references.

Only long papers will be included in the CVPR 2024 proceedings.

Submission Site: https://cmt3.research.microsoft.com/CVPREDGE2024

Submission Deadline: March 22, 2024 (AOE) March 27, 2024 (AOE)

Workshop Date: June 17, 2024 (Full Day)

Workshop Location: Summit 432

Virtual Audience: https://cvpr.thecvf.com/virtual/2024/workshop/23675

We are pleased to announce the accepted papers for the First Workshop on Efficient and On-Device Generation (EDGE) at CVPR 2024. Congratulations to all authors!

Long Papers

• Title: LD-Pruner: Efficient Pruning of Latent Diffusion Models using Task-Agnostic Insights
  Authors: Thibault Castells (Nota Inc.), Hyoung-Kyu Song (Nota Inc.), Bo-Kyeong Kim (Nota Inc.), Shinkook Choi (Nota Inc.)
  [Poster # 36]
• Title: EdgeRelight360: Text-Conditioned 360-Degree HDR Image Generation for Real-Time On-Device Video Portrait Relighting
  Authors: Min-Hui Lin (Qualcomm Technologies Inc.), Mahesh Kumar Krishna Reddy (Qualcomm Technologies Inc), Guillaume Berger (Qualcomm Technologies Inc.), Michel Sarkis (Qualcomm Technologies Inc.), Fatih Porikli (Qualcomm AI Research), Ning Bi (Qualcomm)
  [Poster # 37]

Short Papers

• Title: FoLD: Efficient Fourier-series-based Score Estimation for Langevin Diffusion
  Authors: Siddarth Asokan (Microsoft Research Lab India), Aadithya Srikanth (Indian Institute of Science), Nishanth Shetty (Indian Institute of Science), Chandra Sekhar Seelamantula (IISc Bangalore)
  [Poster # 38]
• Title: EdgeFusion: On-Device Text-to-Image Generation
  Authors: Thibault Castells (Nota Inc.) Hyoung-Kyu Song (Nota Inc.), Tairen Piao (Nota Inc), Shinkook Choi (Nota Inc.), Bo-Kyeong Kim (Nota Inc.), Yim Hanyoung (Samsung Electronics.), Changgwun Lee (Samsung Electronics), Jae Gon Kim (Samsung Electronics), Tae-Ho Kim (Nota, Inc.)
  [Poster # 39]
• Title: Observation-Guided Diffusion Probabilistic Models
  Authors: Junoh Kang (Seoul National University), Jinyoung Choi (Seoul National University), Sungik Choi (LG AI Research), Bohyung Han (Seoul National University)
  [Poster # 40]
• Title: Efficiently Quantize Latent Diffusion Models
  Authors: Yuewei Yang (Meta Platforms Inc.), Jialiang Wang (Meta Platforms Inc), Xiaoliang Dai (Meta Platforms Inc.), Peizhao Zhang (Meta), Hongbo Zhang (Meta)
  [Poster # 41]
• Title: Knowledge Distillation Cross Domain Diffusion Models for Weakly Supervised Defect Detection
  Authors: Yuan-Fu Yang (National Yang Ming Chiao Tung University), Min Sun (NTHU)
  [Poster # 42]
• Title: Bigger is not Always Better: Scaling Properties of Latent Diffusion Models
  Authors: Kangfu Mei (Johns Hopkins University), Zhengzhong Tu (University of Texas at Austin), Mauricio Delbracio (Google Research), Hossein Talebi (Google Inc.), Vishal Patel (Johns Hopkins University), Peyman Milanfar (Google)
  [Poster # 43]
• Title: ECLIPSE: A Resource-Efficient Text-to-Image Prior for Image Generations
  Authors: Maitreya Patel (ASU), Changhoon Kim (Arizona State University), Sheng Cheng (Arizona State University), Chitta Baral (Arizona State University), Yezhou Yang (Arizona State University)
  [Poster # 44]