Copy-paste with self-adaptation: A self-adaptive adjustment method based on copy-paste augmentation

Proceedings Paper Computer Science, Artificial Intelligence

Bootstrapping ViTs: Towards Liberating Vision Transformers from Pre-training

Haofei Zhang et al.

Summary: This paper introduces a method to address the overfitting issue of ViTs on small datasets by incorporating the inductive biases from CNNs and optimizing the agent CNN and ViT through weight sharing. Experimental results demonstrate that the inductive biases help ViTs converge faster and outperform conventional CNNs with fewer parameters.

2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR) (2022)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Towards Open World Object Detection

K. J. Joseph et al.

Summary: Humans have a natural instinct to identify unknown object instances and their curiosity aids in learning about them. The study introduces a novel problem of Open World Object Detection, where a model is tasked to identify unknown objects without explicit supervision and incrementally learn them. The proposed solution ORE shows effectiveness in achieving Open World objectives and reducing confusion in incremental object detection.

2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021 (2021)

Add to Collection

Proceedings Paper Computer Science, Artificial Intelligence

Robust and Accurate Object Detection via Adversarial Learning

Xiangning Chen et al.

Summary: Data augmentation is widely used in training deep image classifiers, but its potential in object detection remains under-explored. This study shows that gains from data augmentations diminish after fine-tuning for object detection, and proposes a model-dependent augmentation method using adversarial examples for better performance. The approach significantly boosts the accuracy and robustness of state-of-the-art object detectors on benchmark datasets.

2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021 (2021)

Add to Collection

Article Computer Science, Artificial Intelligence