Understanding Self-Supervised Pretraining with Part-Aware Representation Learning
This work provides insights into representation learning for computer vision researchers, but it is incremental as it builds on existing self-supervised methods without introducing new techniques.
The paper investigates how self-supervised pretraining methods learn part-aware representations by analyzing contrastive learning as a part-to-whole task and masked image modeling as a part-to-part task, finding that self-supervised models outperform fully-supervised ones in part-level recognition, with combined methods yielding further improvements.
In this paper, we are interested in understanding self-supervised pretraining through studying the capability that self-supervised representation pretraining methods learn part-aware representations. The study is mainly motivated by that random views, used in contrastive learning, and random masked (visible) patches, used in masked image modeling, are often about object parts. We explain that contrastive learning is a part-to-whole task: the projection layer hallucinates the whole object representation from the object part representation learned from the encoder, and that masked image modeling is a part-to-part task: the masked patches of the object are hallucinated from the visible patches. The explanation suggests that the self-supervised pretrained encoder is required to understand the object part. We empirically compare the off-the-shelf encoders pretrained with several representative methods on object-level recognition and part-level recognition. The results show that the fully-supervised model outperforms self-supervised models for object-level recognition, and most self-supervised contrastive learning and masked image modeling methods outperform the fully-supervised method for part-level recognition. It is observed that the combination of contrastive learning and masked image modeling further improves the performance.