Abstract: Methods and systems for training a model include performing two-dimensional object detection on a training image to identify an object. The training image is cropped around the object. A category-level shape reconstruction is generated using a neural radiance field model. A normalized coordinate model is trained using the training image and ground truth information from the category-level shape reconstruction.

Abstract: Methods and systems for training a model include training a size estimation model to generate an estimated object size using a training dataset with differing levels of annotation. Two-dimensional object detection is performed on a training image to identify an object. The training image is cropped around the object. A category-level shape reconstruction is generated using a neural radiance field model. A normalized coordinate model is trained using the training image and ground truth information from the category-level shape reconstruction.

Abstract: Techniques are described for using computing devices to perform automated operations for determining the acquisition location of an image using an analysis of the image's visual contents. In at least some situations, images to be analyzed include panorama images acquired at acquisition locations in an interior of a multi-room building, and the determined acquisition location information includes a location on a floor plan of the building and in some cases orientation direction information—in at least some such situations, the acquisition location determination is performed without having or using information from any distance-measuring devices about distances from an image's acquisition location to objects in the surrounding building. The acquisition location information may be used in various automated manners, including for controlling navigation of devices (e.g., autonomous vehicles), for display on one or more client devices in corresponding graphical user interfaces, etc.