8th Workshop on Efficient Deep Learning for Computer Vision

Workshop at CVPR 2025

Recent advances in deep learning have provided significant improvements in the ability to understand visual content. Mainstream computer vision research often gives little consideration to computation time, resource requirements, and constraints such as energy consumption, memory footprint and model size, or carbon emission. As computer vision is increasingly deployed in resource-limited systems (such as mobile phones, wearable devices, Internet of Things, autonomous aerial vehicles), addressing all of these metrics becomes essential. In this year’s ECV workshop, the topics of interest include but are not limited to the following:

Topics:

  • Mobile Generative vs Discriminative models
  • Augmented and Virtual Reality for computer vision, such as image and video synthesis, object tracking, object recognition, style transfer
  • Integration of large language models (LLM) and computer vision, such as language-guided synthesis of images or videos; as well as image understanding via Vision-Language Models.
  • Applications of computer vision running on mobile or wearable devices
  • Training on mobile devices
  • User experience of computer vision
  • Efficient Foundation Models
    • Compact and efficient architectures for foundation models, such as Vision-Language Models, Generative Adversarial Networks (GANs), Autoregressive models, Denoising Diffusion Probabilistic Models (DDPMs) and Multi-Modal Language models.
    • Efficient generative architecture search algorithms for generative model tasks (domain adaptation, style transfer,Latency-aware loss design for efficient model training)
    • Tunable efficiency-accuracy tradeoffs
  • Efficient Neural Architecture, Compression, Quantization and Hardware Acceleration
    • Compact and efficient neural network architecture design
    • Efficient neural architecture search algorithms for different vision tasks (detection, segmentation, generative models, etc.)
    • Hardware accelerators to support computer vision
    • Hardware-software codesign for computer vision
  • Data-Efficient Learning
    • Methods, algorithms, systems that improves data efficiency of deep learning
    • Different learning modalities (e.g., supervised, self-supervised, reinforcement) that can improve efficiency for computer vision
    • Zero-shot, few-shots learning, language-supervised learning
    • Open-set recognition, detection, and segmentation
  • Efficient 3D models
    • Efficient models for 3D understanding
    • Efficient neural rendering