In the field of Class Incremental Object Detection (CIOD), creating models that can continuously learn like humans is a major challenge. Pseudo-labeling methods, although initially powerful, struggle with multi-
Class-Incremental Learning with CLIP: Adaptive Representation Adjustment and Parameter Fusion (ECCV24)[paper][code] Bridge Past and Future: Overcoming Information Asymmetry in Incremental Object Detection (ECCV24)[paper][code] Confidence Self-Calibration for Multi-Label Class-Incremental Learning (ECCV24...
Learning with Fantasy: Semantic-Aware Virtual Contrastive Constraint for Few-Shot Class-Incremental Learning (CVPR2023)[paper][code] Foundation Model Drives Weakly Incremental Learning for Semantic Segmentation (CVPR2023)[paper] Continual Detection Transformer for Incremental Object Detection (CVPR2023)[paper...
For each input video, we first extract multi-grained features with the image encoder of a CLIP model, then select corresponding prompts from predefined prompt pools. Finally, the prompts are prepended to the video and text embedding feature respec- tively to instruct the model ...
All experiments used the CLIP-based ResNet as the feature extractor. The miniImageNet results are listed in Table 6. We observe a decrease in accuracy with the few-shot setting when compared to the full-labeled design, but this difference shrinks after the last incremental session. The result...
Firstly, by first introducing CLIP, a foundation deep learning model that has not yet been applied in animal fields, the powerful recognition capability with few training resources is exploited with an additional shallow network. Secondly, inspired by the single-image recognition abilities of zoologist...
Learning with Fantasy: Semantic-Aware Virtual Contrastive Constraint for Few-Shot Class-Incremental Learning (CVPR2023)[paper][code] Foundation Model Drives Weakly Incremental Learning for Semantic Segmentation (CVPR2023)[paper] Continual Detection Transformer for Incremental Object Detection (CVPR2023)[paper...