Abstract: Systems and techniques are provided for generating depth information. For example, a process can include obtaining a first feature volume including visual features corresponding to each respective frame included in a first set of frames. A first query generator network can generate reconstruction features associated with a reconstructed feature volume corresponding to the first feature volume. Based on the first feature volume, a second query generator network can generate motion features associated with predicted future motion corresponding to the first feature volume. An initial depth prediction can be generated for each respective frame based on cross-attention between features of a depth prediction decoder, the reconstruction features, and the motion features. A refined depth prediction can be generated for each respective based on cross-attention between the initial depth prediction, the reconstruction features, and the motion features.

Abstract: Systems and techniques described herein relate to generating an inter-frame from a first and second frame. An apparatus includes a memory storing a first frame and a second frame; and a processor coupled to the memory and configured to: estimate at least one optical flow between the first frame and the second frame; generate, based on the at least one optical flow, at least one occlusion mask; generate, based on the at least one optical flow and the at least one occlusion mask, at least one weighting mask; generate, based on the at least one optical flow and the at least one weighting mask, at least one inter-frame optical flow; generate, based on the at least one inter-frame optical flow and at least one of the first frame or the second frame, at least one warped frame; and generate, based on the at least one warped frame, an inter-frame.

Abstract: A method for training a diffusion model includes randomly selecting, for each iteration of a step distillation training process, a teacher model of a group of teacher models. The method also includes applying, at each iteration, a clipped input space within step distillation of the randomly selected teacher model. The method further includes updating, at each iteration, parameters of the diffusion model based on guidance from the randomly selected teacher model.

Abstract: A method for training a control neural network includes initializing a baseline diffusion model for training the control neural network, each stage of a control neural network training pipeline corresponding to an element of the baseline diffusion model. The method also includes training, the control neural network, in a stage-wise manner, each stage of the control neural network training pipeline receiving an input from a previous stage of the control neural network training pipeline and the corresponding element of the diffusion model.

Abstract: A processor-implemented method includes receiving a text-semantic input at a first stage of a neural network, including a first convolutional block and no attention layers. The method receives, at a second stage, a first output from the first stage. The second stage comprises a first down sampling block including a first attention layer and a second convolutional block. The method receives, at a third stage, a second output from the second stage. The third stage comprises a first up sampling block including a second attention layer and a first set of convolutional blocks. The method receives, at a fourth stage, the first output from the first stage and a third output from the third stage. The fourth stage comprises a second up sampling block including no attention layers and a second set of convolutional blocks. The method generates an image at the fourth stage, based on the text-semantic input.

Abstract: A method for training a diffusion model includes compressing the diffusion model by removing at least one of: one or more model parameters or one or more giga multiply-accumulate operations (GMACs). The method also includes performing guidance conditioning to train the compressed diffusion model, the guidance conditioning combining a conditional output and an unconditional output from respective teacher models. The method further includes performing, after the guidance conditioning, step distillation on the compressed diffusion model.

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for improved machine learning. During a first iteration of processing data using a denoising backbone of a diffusion machine learning model, a first latent tensor is generated using a lower resolution block of the denoising backbone, and a first feature tensor is generated based on processing the first latent tensor using a higher resolution block of the denoising backbone, the higher resolution block using a higher resolution than the lower resolution block. A second latent tensor is generated based on processing the first latent tensor using an adapter block of the denoising backbone. During a second iteration of processing the data using the denoising backbone, a second feature tensor is generated based on processing the second latent tensor using the higher resolution block.

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for improved machine learning. During a first iteration of processing data using a first denoising backbone of a teacher diffusion machine learning model, a first latent tensor is generated using a lower resolution block of the first denoising backbone. During a first iteration of processing data using a second denoising backbone of a student diffusion machine learning model, a second latent tensor is generated using an adapter block of the second denoising backbone. A loss is generated based on the first and second latent tensors, and one or more parameters of the adapter block are updated based on the loss.

Abstract: Techniques and systems are provided for image processing. For instance, a process can include correlating a first set of features from a first viewpoint with a second set of features from a second viewpoint at a first time period to generate a first disparity cost volume; correlating a third set of features from the first viewpoint at a second time period with the first set of features to generate a first optical flow cost volume; gating the first disparity cost volume to generate first intermediate disparity information; gating the first optical flow cost volume to generate first intermediate optical flow information; correlating the first set of features, the second set of features, and the first intermediate optical flow information to generate disparity information for output; and correlating the third set of features, the first set of features, and the first intermediate disparity information to generate optical flow information for output.

Abstract: A method of processing image content includes constructing a first graph representation having a first level of point sparsity from a first point cloud data, and performing diffusion-based upsampling on the first graph representation to generate a second graph representation having a second level of point sparsity. Performing diffusion-based upsampling includes inputting the first graph representation into a diffusion-based trained model to generate a first intermediate graph representation having a first intermediate level of point sparsity, inputting the first intermediate graph representation into the diffusion-based trained model to generate a second intermediate graph representation having a second intermediate level of point sparsity, and generating the second graph representation based on at least on the second intermediate graph representation.

Abstract: Systems and techniques are provided for processing image data. According to some aspects, a computing device can generate a gradient (e.g., a classifier gradient using a trained classifier) associated with a current sample. The computing device can combine the gradient with an iterative model estimated score function or data associated with the current sample to generate a score function estimate. The computing device can predict, using the diffusion machine learning model and based on the score function estimate, a new sample.

Abstract: Systems and techniques are provided for generating disparity information from two or more images. For example, a process can include obtaining first disparity information corresponding to a pair of images, the pair of images including a first image of a scene and a second image of the scene. The process can include obtaining confidence information associated with the first disparity information. The process can include processing, using a machine learning network, the first disparity information and the confidence information to generate second disparity information corresponding to the pair of images. The process can include combining, based on the confidence information, the first disparity information with the second disparity information to generate a refined disparity map corresponding to the pair of images.

Abstract: Disclosed are systems and techniques for capturing images (e.g., using a monocular image sensor) and detecting depth information. According to some aspects, a computing system or device can generate a feature representation of a current image and update accumulated feature information for storage in a memory based on a feature representation of a previous image and optical flow information of the previous image. The accumulated feature information can include accumulated image feature information associated with a plurality of previous images and accumulated optical flow information associated of the plurality of previous images. The computing system or device can obtain information associated with relative motion of the current image based on the accumulated feature information and the feature representation of the current image.

Abstract: A computer-implemented method includes receiving a first input. The first input is interpolated based on a first shift along a first dimension and a second shift along a second dimension. A first output is generated based on the interpolated first input. The first output corresponds to a vectorized bilinear shift of the first input for use in place of grid sampling algorithms.

Abstract: Techniques and systems are provided for generating one or more segmentations masks. For instance, a process may include generating a delta image based on a difference between a current image and a prior image. The process may further include processing, using a transform operation, the delta image and features representing the prior image to generate a transformed feature representation of the prior image. The process may include combining the transformed feature representation of the prior image with features representing the current image to generate a combined feature representation of the current image. The process may further include generating, based on the combined feature representation of the current image, a segmentation mask for the current image.

Abstract: A computer-implemented method includes generating a first augmented frame by combining a first image and a first frame of a first frame pair. The computer-implemented method also includes generating, via an optical flow estimation model, a first flow estimation based on a second frame of the first frame pair and the first augmented frame. The computer-implemented method further includes updating one or both of parameters or weights of the optical flow estimation model based on a first loss between the first flow estimation and a training target.

Abstract: Certain aspects of the present disclosure provide techniques and apparatus for regenerative learning to enhance dense predictions. In one example method, an input image is accessed. A dense prediction output is generated based on the input image using a dense prediction machine learning (ML) model, and a regenerated version of the input image is generated. A first loss is generated based on the input image and a corresponding ground truth dense prediction, and a second loss is generated based on the regenerated version of the input image. One or more parameters of the dense prediction ML model are updated based on the first and second losses.

Abstract: A processor-implemented method comprises processing a single level cost volume across multiple processing stages by varying a receptive field across each of the processing stages. The method also includes performing a learning-based correspondence estimation task based on the processing. The varying may include processing a different resolution of the cost volume at each processing stage while maintaining a same neighborhood sampling radius. The resolution may increase from a first processing stage to a later processing stage. The varying may also include varying a neighborhood sampling radius at each of the processing stages while maintaining a same resolution. The task may be optical flow estimation or stereo estimation.

Abstract: Systems and techniques are provided for processing one or more images. For instance, according to some aspects of the disclosure, a method may include obtaining an unlabeled image and generating at least one transformed image based on the unlabeled image. The method may include processing the unlabeled image using a pre-trained semantic segmentation model to generate a first segmentation output. The method may further include processing the at least one transformed image using the pre-trained semantic segmentation model to generate at least a second segmentation output. The method may include fine-tuning, based on the first segmentation output and at least the second segmentation output, one or more parameters of the pre-trained semantic segmentation model.

Abstract: Systems and techniques are provided for processing data (e.g., image data). For instance, according to some aspects of the disclosure, a method may include receiving, at a transformer of a machine learning system, learnable queries, keys, and values obtained from a feature map of a segmentation model of the machine learning system. The method may further include learning, via the transformer, a mapping between an unsupervised output and a supervised output of the segmentation model based on the feature map.