CNN-JEPA: Self-Supervised Pretraining Convolutional Neural Networks Using Joint Embedding Predictive Architecture
This provides a simpler and more efficient self-supervised pretraining method for CNNs, addressing a domain-specific bottleneck in computer vision.
The paper tackles adapting self-supervised learning methods to Convolutional Neural Networks by introducing CNN-JEPA, which achieves 73.3% linear top-1 accuracy on ImageNet-100 and reduces training time by 17-35% compared to other CNN-based SSL methods.
Self-supervised learning (SSL) has become an important approach in pretraining large neural networks, enabling unprecedented scaling of model and dataset sizes. While recent advances like I-JEPA have shown promising results for Vision Transformers, adapting such methods to Convolutional Neural Networks (CNNs) presents unique challenges. In this paper, we introduce CNN-JEPA, a novel SSL method that successfully applies the joint embedding predictive architecture approach to CNNs. Our method incorporates a sparse CNN encoder to handle masked inputs, a fully convolutional predictor using depthwise separable convolutions, and an improved masking strategy. We demonstrate that CNN-JEPA outperforms I-JEPA with ViT architectures on ImageNet-100, achieving a 73.3% linear top-1 accuracy using a standard ResNet-50 encoder. Compared to other CNN-based SSL methods, CNN-JEPA requires 17-35% less training time for the same number of epochs and approaches the linear and k-NN top-1 accuracies of BYOL, SimCLR, and VICReg. Our approach offers a simpler, more efficient alternative to existing SSL methods for CNNs, requiring minimal augmentations and no separate projector network.