Search Results for "pixels."

JEPA (Joint-Embedding Predictive Architecture) captures the idea of predicting missing high-level representations rather than reconstructing pixels, aiming for robust, scalable self-supervised learning. A context encoder ingests visible regions and predicts target embeddings for masked regions produced by a separate target encoder, avoiding low-level reconstruction losses that can overfit to texture. This makes learning focus on semantics and structure, yielding features that transfer well with simple linear probes and minimal fine-tuning. ...

VJEPA2 is a next-generation self-supervised learning framework for video that extends the “predict in representation space” idea from i-JEPA to the temporal domain. Instead of reconstructing pixels, it predicts the missing high-level embeddings of masked space-time regions using a context encoder and a slowly updated target encoder. This objective encourages the model to learn semantics, motion, and long-range structure without the shortcuts that pixel-level losses can invite. The architecture is designed to scale: spatiotemporal ViT backbones, flexible masking schedules, and efficient sampling let it train on long clips while remaining stable. ...

Consistent Depth is a research project developed by Facebook Research that presents an algorithm for reconstructing dense and geometrically consistent depth information for all pixels in a monocular video. The system builds upon traditional structure-from-motion (SfM) techniques to provide geometric constraints while integrating a convolutional neural network trained for single-image depth estimation. During inference, the model fine-tunes itself to align with the geometric constraints of a specific input video, ensuring stable and realistic depth maps even in less-constrained regions. ...