Dense self-supervised learning has shown great promise for learning pixel- and patch-level representations, but extending it to videos remains challenging due to the complexity of motion dynamics. Existing approaches struggle as they rely on static augmentations that fail under object deformations, occlusions, and camera movement, leading to inconsistent feature learning over time. We propose a motion-guided self-supervised learning framework that clusters dense point tracks to learn spatiotemporally consistent representations. By leveraging an off-the-shelf point tracker, we extract long-range motion trajectories and optimize feature clustering through a momentum-encoder-based optimal transport mechanism. To ensure temporal coherence, we propagate cluster assignments along tracked points, enforcing feature consistency across views despite viewpoint changes. Integrating motion as an implicit supervisory signal, our method learns representations that generalize across frames, improving robustness in dynamic scenes and challenging occlusion scenarios. By initializing from strong image-pretrained models and leveraging video data for training, we improve state-of-the-art by 1% to 6% on six image and video datasets and four evaluation benchmarks. The implementation is publicly available at our GitHub repository: https://github.com/SMSD75/MoSiC/tree/main

This work aims to improve the efficiency of vision transformers (ViTs). While ViTs use computationally expensive self-attention operations in every layer, we identify that these operations are highly correlated across layers -- a key redundancy that causes unnecessary computations. Based on this observation, we propose SkipAT a method to reuse self-attention computation from preceding layers to approximate attention at one or more subsequent layers. To ensure that reusing self-attention blocks across layers does not degrade the performance, we introduce a simple parametric function, which outperforms the baseline transformer's performance while running computationally faster. We show that SkipAT is agnostic to transformer architecture and is effective in image classification, semantic segmentation on ADE20K, image denoising on SIDD, and video denoising on DAVIS. We achieve improved throughput at the same-or-higher accuracy levels in all these tasks.

Self-supervised learning has unlocked the potential of scaling up pretraining to billions of images, since annotation is unnecessary. But are we making the best use of data? How more economical can we be? In this work, we attempt to answer this question by making two contributions. First, we investigate first-person videos and introduce a ``Walking Tours'' dataset. These videos are high-resolution, hours-long, captured in a single uninterrupted take, depicting a large number of objects and actions with natural scene transitions. They are unlabeled and uncurated, thus realistic for self-supervision and comparable with human learning. Second, we introduce a novel self-supervised image pretraining method tailored for learning from continuous videos. Existing methods typically adapt image-based pretraining approaches to incorporate more frames. Instead, we advocate a ``tracking to learn to recognize'' approach. Our method called DoRA, leads to attention maps that **D**isc**O**ver and t**RA**ck objects over time in an end-to-end manner, using transformer cross-attention. We derive multiple views from the tracks and use them in a classical self-supervised distillation loss. Using our novel approach, a single Walking Tours video remarkably becomes a strong competitor to ImageNet for several image and video downstream tasks.

Mixup refers to interpolation-based data augmentation, originally motivated as a way to go beyond empirical risk minimization (ERM). Its extensions mostly focus on the definition of interpolation and the space (input or feature) where it takes place, while the augmentation process itself is less studied. In most methods, the number of generated examples is limited to the mini-batch size and the number of examples being interpolated is limited to two (pairs), in the input space. We make progress in this direction by introducing MultiMix, which generates an arbitrarily large number of interpolated examples beyond the mini-batch size and interpolates the entire mini-batch in the embedding space. Effectively, we sample on the entire convex hull of the mini-batch rather than along linear segments between pairs of examples. On sequence data, we further extend to Dense MultiMix. We densely interpolate features and target labels at each spatial location and also apply the loss densely. To mitigate the lack of dense labels, we inherit labels from examples and weight interpolation factors by attention as a measure of confidence. Overall, we increase the number of loss terms per mini-batch by orders of magnitude at little additional cost. This is only possible because of interpolating in the embedding space. We empirically show that our solutions yield significant improvement over state-of-the-art mixup methods on four different benchmarks, despite interpolation being only linear. By analyzing the embedding space, we show that the classes are more tightly clustered and uniformly spread over the embedding space, thereby explaining the improved behavior.

Mixup is a powerful data augmentation method that in-terpolates between two or more examples in the input or feature space and between the corresponding target labels. However, how to best interpolate images is not well defined. Recent mixup methods overlay or cut-and-paste two or more objects into one image, which needs care in selecting regions. Mixup has also been connected to autoencoders, because often autoencoders generate an image that continuously deforms into another. However, such images are typically of low quality. In this work, we revisit mixup from the deformation perspective and introduce AligtiMixup, where we geometrically align two images in the feature space. The correspondences allow us to interpolate between two sets of features, while keeping the locations of one set. Interestingly, this retains mostly the geometry or pose of one image and the appearance or texture of the other. We also show that an autoencoder can still improve representation learning under mixup, without the classifier ever seeing decoded images. AlignMixup outperforms state-of-the-art mixup methods on five different benchmarks. Code available at https://github.com/shashankvkt/AlignMixup_CVPR22.git

The detection of targets in infra-red imagery is a challenging problem which involves locating small targets in heavily cluttered environments while maintaining a low false alarm rate. We propose a network that optimizes a “target to clutter ratio” (TCR) metric defined as the ratio of the output energies produced by the network in response to targets and clutter. We show that for target detection, it is advantageous to analytically derive the first layer of a CNN to maximize the TCR metric, and then train the rest of the network to optimize the same cost function. We evaluate the performance of the resulting network using a public domain MWIR data set released by the US Army's Night Vision Laboratories, and compare it to the state-of-the-art detectors such as Faster RCNN and Yolo-v3. Referred to as the TCRNet, the proposed network demonstrates state of the art results with greater than 30% improvement in probability of detection while reducing the false alarm rate by more than a factor of 2 when compared to these leading methods. Ablation studies also show that the proposed approach and metric are superior to learning the entire network from scratch, or using conventional regression metrics such as the mean square error (MSE).

Recently, near-infrared (NIR) images are increasingly being captured for recognizing faces in low-light/night-time conditions. Matching these images against the controlled high-resolution visible facial images usually present in the database is a challenging task. In surveillance scenarios, the NIR images can have very low resolution and also non-frontal pose which makes the problem even more challenging. In this paper, we propose an orthogonal dictionary alignment approach for addressing this problem. We also propose a re-ranking approach to further improve the recognition performance for each probe by combining the rank list given by the proposed algorithm with that given by another complementary feature/algorithm. Finally, we have also collected our own database Heterogeneous face recognition across Pose and Resolution (HPR) that has facial images captured from two surveillance quality NIR cameras and one high-resolution visible camera, with significant variations in head pose and resolution. Extensive experiments on the modified CASIA NIR-VIS 2.0 database, the Surveillance Camera face database, and our HPR database show the effectiveness of the proposed approaches and the collected database.