ICCV 2023: Selected Posters

Vision-Language Models

• Delving into CLIP latent space for Video Anomaly Detection and Recognition



• SMAUG: Sparse Masked Autoencoder for Efficient Video-Language Pre-training



• Distilling Large Vision-Language Model with Out-of-Distribution Generalizability



• Prompt Switch: Efficient CLIP Adaptation for Text-Video Retrieval



• Black Box Few-Shot Adaptation for Vision-Language Models



• TinyCLIP: CLIP Distillation via Affinity Mimicking and Weight Inheritance



• BlendShift: Adaptive Neighbour Correction and Replacement for Efficient Neighbour Contrastive Learning

Vision Models

• SeiT: Storage-Efficient Vision Training with Tokens Using 1% of Pixel Storage



• Rethinking Vision Transformer from the View of Path Ensemble



• XiNet: Efficient Neural Networks for tinyML



• Convolutional Networks with Oriented 1D Kernels



• FLatten Transformer: Vision Transformer using Focused Linear Attention



• Rethinking Mobile Block for Efficient Attention-based Models



• BiViT: Extremely Compressed Binary Vision Transformers

Egocentric

• COPliOT: Human-Environment Collision Prediction and Localization from Multi-view Egocentric Videos



• EGO-ONLY: EGOCENTRIC ACTION DETECTION WITHOUT EXOCENTRIC TRANSFERRING

Video

• MiniROAD: Minimal RNN Framework for Online Action Detection



• Spatio-temporal Prompting Network for Robust Video Feature Extraction



• Label-Efficient Online Continual Object Detection in Streaming Video



• Efficient Video Prediction via Sparsely Conditioned Flow Matching

Anonymization

• PRIFACE: A Novel Pipeline for Enhancing and Evaluating Face Anonymization