Toyota Research Institute, Inc.

Toyota Jidosha Kabushiki Kaisha

Siemens Aktiengesellschaft

Leland Stanford Junior University

Massachusetts Institute of Technology

Sergey Zakharov

Rares A Ambrus

Vitor Guizilini

Adrien David Gaidon

Slobodan Ilic

Ivan Shugurov

Andreas Hutter

Jie Li

Dennis Park

Fredo Durand

Tolga Birdal

Roman Kaskman

Jiajun Wu

Joshua Tenenbaum

Wadim Kehl

Dian Chen

Vincent Sitzmann

Davis Rempe

Colton Stearns

A method for multiple object tracking includes receiving, with a computing device, a point cloud dataset, detecting one or more objects in the point cloud dataset, each of the detected one or more objects defined by points of the point cloud dataset and a bounding box, querying one or more historical tracklets for historical tracklet states corresponding to each of the one or more detected objects, implementing a 4D encoding backbone comprising two branches: a first branch configured to compute per-point features for each of the one or more objects and the corresponding historical tracklet states, and a second branch configured to obtain 4D point features, concatenating the per-point features and the 4D point features, and predicting, with a decoder receiving the concatenated per-point features, current tracklet states for each of the one or more objects.

Systems and methods for multi-object tracking

A method for improving 3D object detection via object-level augmentations is described. The method includes recognizing, using an image recognition model of a differentiable data generation pipeline, an object in an image of a scene. The method also includes generating, using a 3D reconstruction model, a 3D reconstruction of the scene from the image including the recognized object. The method further includes manipulating, using an object level augmentation model, a random property of the object by a random magnitude at an object level to determine a set of properties and a set of magnitudes of an object manipulation that maximizes a loss function of the image recognition model. The method also includes training a downstream task network based on a set of training data generated based on the set of properties and the set of magnitudes of the object manipulation, such that the loss function is minimized.

Adversarial object-aware neural scene rendering for 3D object detection

Various embodiments of the teachings herein include a computer implemented pose estimation method for providing poses of objects of interest in a scene. The scene comprises a visual representation of the objects of interest in an environment. The method comprising: conducting for each one of the objects of interest a pose estimation; determining edge data of the object of interest from the visual representation representing the edges of the respective object of interest; determining keypoints of the respective object of interest by a previously trained artificial neural keypoint detection network, wherein the artificial neural keypoint detection network utilizes the determined edge data of the respective object of interest Oi as input and provides the keypoints of the respective object of interest as output; and estimating the pose of the respective object of interest based on the respective object's keypoints provided by the artificial neural keypoint detection network.

Pose estimation method and apparatus

System, methods, and other embodiments described herein relate to an improved approach to training a depth model for monocular depth estimation by warping depth features prior to decoding. In one embodiment, a method includes encoding, using an encoder of a depth model, a source image into depth features of a scene depicted by the source image. The method includes warping the depth features into warped features of a target frame of a target image associated with the source image. The method includes decoding, using a decoder of the depth model, the warped features into a depth map. The method includes training the depth model according to a loss derived from the depth map.

Warping depth features for depth estimation

System, methods, and other embodiments described herein relate to an improved approach to training a depth model to derive depth estimates from monocular images using histograms to assess photometric losses. In one embodiment, a method includes determining loss values according to a photometric loss function. The loss values are associated with a depth map derived from an input image that is a monocular image. The method includes generating histograms for the loss values corresponding to different regions of a target image. The method includes, responsive to identifying erroneous values of the loss values, masking the erroneous values to avoid considering the erroneous values during training of the depth model.

Using histograms for self-supervised depth estimation

System, methods, and other embodiments described herein relate to an improved approach to training a depth model to derive depth estimates from monocular images using cost volumes. In one embodiment, a method includes predicting, using a depth model, depth values from at least one input image that is a monocular image. The method includes generating a cost volume by sampling the depth values corresponding to bins of the cost volume. The method includes determining loss values for the bins of the cost volume. The method includes training the depth model according to the loss values of the cost volume.

Photometric cost volumes for self-supervised depth estimation

System, methods, and other embodiments described herein relate to training a depth model for monocular depth estimation using photometric loss masks derived from motion estimates of dynamic objects. In one embodiment, a method includes generating depth maps from images of an environment. The method includes determining motion of points within the depth maps. The method includes associating the points between the depth maps to identify an object according to a correlation of the motion for a first cluster of the points with a second cluster of the points. The method includes providing the depth maps and the object as an electronic output.

Self-supervised monocular depth estimation via rigid-motion embeddings

A system for producing a depth map can include a processor and a memory. The memory can store a neural network. The neural network can include an encoding portion module, a multi-frame feature matching portion module, and a decoding portion module. The encoding portion module can include instructions that, when executed by the processor, cause the processor to encode an image to produce single-frame features. The multi-frame feature matching portion module can include instructions that, when executed by the processor, cause the processor to process the single-frame features to produce information. The decoding portion module can include instructions that, when executed by the processor, cause the processor to decode the information to produce the depth map. A first training dataset, used to train the multi-frame feature matching portion module, can be different from a second training dataset used to train the encoding portion module and the decoding portion module.

Producing a depth map from a monocular two-dimensional image

In accordance with one embodiment of the present disclosure, a method includes receiving an input image having an object and a background, intrinsically decomposing the object and the background into an input image data having a set of features, augmenting the input image data with a 2.5D differentiable renderer for each feature of the set of features to create a set of augmented images, and compiling the input image and the set of augmented images into a training data set for training a downstream task network.

Guided domain randomization via differentiable dataset rendering

In accordance with one embodiment of the present disclosure, a method includes receiving a set of images, each image depicting a view of a scene, generating sparse depth data from each image of the set of images, training a monocular depth estimation model with the sparse depth data, generating, with the trained monocular depth estimation model, depth data and uncertainty data for each image, training a NeRF model with the set of images, wherein the training is constrained by the depth data and uncertainty data, and rendering, with the trained NeRF model, a new image having a new view of the scene.

Systems and methods for neural implicit scene representation with dense, uncertainty-aware monocular depth constraints

A method and a system for object detection and pose estimation within an input image. A 6-degree-of-freedom object detection and pose estimation is performed using a trained encoder-decoder convolutional artificial neural network including an encoder head, an ID mask decoder head, a first correspondence color channel decoder head and a second correspondence color channel decoder head. The ID mask decoder head creates an ID mask for identifying objects, and the color channel decoder heads are used to create a 2D-to-3D-correspondence map. For at least one object identified by the ID mask, a pose estimation based on the generated 2D-to-3D-correspondence map and on a pre-generated bijective association of points of the object with unique value combinations in the first and the second correspondence color channels is generated.

Dense 6-DoF pose object detector

Systems and methods described herein relate to reconstructing a scene in three dimensions from a two-dimensional image. One embodiment processes an image using a detection transformer to detect an object in the scene and to generate a NOCS map of the object and a background depth map; uses MLPs to relate the object to a differentiable database of object priors (PriorDB); recovers, from the NOCS map, a partial 3D object shape; estimates an initial object pose; fits a PriorDB object prior to align in geometry and appearance with the partial 3D shape to produce a complete shape and refines the initial pose estimate; generates an editable and re-renderable 3D scene reconstruction based, at least in part, on the complete shape, the refined pose estimate, and the depth map; and controls the operation of a robot based, at least in part, on the editable and re-renderable 3D scene reconstruction.

Growing community of inventors

San Francisco, CA, United States of America

Sergey Zakharov