A storage medium storing a training processing program that causes at least one computer to execute a process that includes acquiring a deviation degree of a feature in a training dataset, by using a determination model, the training dataset being unlabeled; selecting one or more pieces of data included in the training dataset based on the deviation degree; outputting the selected one or more pieces of data or related data related to the selected one or more pieces of data; receiving an input of a determination result by a user for the one or more pieces of data; and determining an adjustment standard used to adjust a feature of each piece of the data included in the training dataset based on the received determination result, wherein when determination target data is determined by the determination model, a feature of the determination target data is adjusted based on the adjustment standard.

A method for automatically estimating cellularity in a digital pathology slide image includes: extracting patches of interest from the digital pathology slide image; operating on each patch using a trained first deep convolutional neural network (DCNN) to classify that patch as either normal, having an estimated cellularity of 0%, or suspect, having a cellularity roughly estimated to be greater than 0%; operating on each suspect patch using a second DCNN, trained using a deep ordinal regression model, to determine an estimated cellularity score for that suspect patch; and combining the estimated cellularity scores of the patches of interest to provide an estimated cellularity for the digital pathology slide image at a patch-by-patch level.

A system for training a neural network for action recognition based on unlabeled action sequences includes a first neural network (NN1) and a second neural network (NN2). A first updating module is arranged to update parameters of NN1 to minimize a difference between representation data generated by NN1 and representation data generated by NN2. A second updating module is arranged to update parameters of NN2 as a function of the parameters of NN1. An augmentation module includes first and second sub-modules and is configured to include augmented versions of incoming action sequences in first and second input data. The first and second sub-modules are configured to apply at least partly different augmentation to the incoming action sequences. After NN1 and NN2 have been operated on one or more instances of the first and second input data, NN1 comprises a parameter definition of a pre-trained neural network.

Methods and systems for training a model labeling two or more organic structures within an image. One method includes receiving a set of training images. The set of training images including a first plurality of images and a second plurality of images. Each of the first plurality of images including a label for each of the two or more organic structures and each of the second plurality of images including a label for only a subset of the two or more organic structures. The method further includes training the model using the first plurality of images, the second plurality of images, and a label merging function mapping a label from the first plurality of images to a label included in the second plurality of images.

Some embodiments relate to generating three dimensional virtual representations of a building construction structure based on two-dimensional real-world construction plans, such as architectural plans or building plans. Some embodiments further produce autonomous, near real-time, and highly accurate and comprehensive building take-offs, complete construction detailing or estimates, detailed bill of materials, plan analysis (including detection of a number of non-standardized objects, such as doors or windows), as well as transforming 2D drawings into 3D and/or providing Building Information Modeling (BIM). The two dimensional real-world architectural plan can include multivariate non-standardized architectural symbols, which define numerous objects including trees, bathrooms, doors, stairs, windows, and floor finishes, lines, including solid, hollow, dashed and dotted lines, which define features including internal or external walls, windows, doors, stairs, property boundaries, easements, footpaths, rooflines, driveways, rights of way, paving stones, landscaping, water, power, drainage, and dimensions, shading, and patterns which define materials and areas on the two dimensional real-world architectural plan, and text which indicate the purposes of the rooms, dimensions, features, construction methods, and regulatory standards.

A processing apparatus includes a collection module and a training module, the training module includes a backbone network and a region proposal network (RPN) layer, the backbone network is connected to the RPN layer, and the RPN layer includes a class activation map (CAM) unit. The collection module is configured to obtain an image, where the image includes an image with an instance-level label and an image with an image-level label. The backbone network is used to output a feature map of the image based on the image obtained by the collection module.

Embodiments provide electronic methods and systems for improving edge case classifications. The method performed by a server system includes accessing an input sample dataset including first labeled training data associated with a first class, and second labeled training data associated with a second class, from a database. Method includes executing training of a first autoencoder and a second autoencoder based on the first and second labeled training data, respectively. Method includes providing the first and second labeled training data along with unlabeled training data accessed from the database to the first and second autoencoders. Method includes calculating a common loss function based on a combination of a first reconstruction error associated with the first autoencoder and a second reconstruction error associated with the second autoencoder. Method includes fine-tuning the first autoencoder and the second autoencoder based on the common loss function.

Systems and methods for obstacle detection are provided. The system aligns image level features between a source domain and a target domain based on an adversarial learning process while training a domain discriminator. The target domain includes one or more road scenes having obstacles. The system selects, using the domain discriminator, unlabeled samples from the target domain that are far away from existing annotated samples from the target domain. The system selects, based on a prediction score of each of the unlabeled samples, samples with lower prediction scores. The system annotates the samples with the lower prediction scores.

Systems and methods of the present disclosure provide an improved approach for open-set instance segmentation by identifying both known and unknown instances in an environment. For example, a method can include receiving sensor point cloud input data including a plurality of three-dimensional points. The method can include determining a feature embedding and at least one of an instance embedding, class embedding, and/or background embedding for each of the plurality of three-dimensional points. The method can include determining a first subset of points associated with one or more known instances within the environment based on the class embedding and the background embedding associated with each point in the plurality of points. The method can include determining a second subset of points associated with one or more unknown instances within the environment based on the first subset of points. The method can include segmenting the input data into known and unknown instances.

A computer-implemented method of learning sensory media association includes receiving a first type of nontext input and a second type of nontext input; encoding and decoding the first type of nontext input using a first autoencoder having a first convolutional neural network, and the second type of nontext input using a second autoencoder having a second convolutional neural network; bridging first autoencoder representations and second autoencoder representations by a deep neural network that learns mappings between the first autoencoder representations associated with a first modality and the second autoencoder representations associated with a second modality; and based on the encoding, decoding, and the bridging, generating a first type of nontext output and a second type of nontext output based on the first type of nontext input or the second type of nontext input in either the first modality or the second modality.