Character And Pattern Recognition TRIZ Cases

A tomosynthesis display system renders volumes of interest using slice, slab, or 3D views.

A self-checkout system segments fraud detection into independent modules to assign cumulative points based on customer behavior patterns.

A weakly supervised video activity detection method uses iterative learning to refine class activation sequences and fuse semantic features with video features.

A training method uses Markov chain cropping probabilities to optimize image processing networks.

An augmented reality display system uses a mirror to create virtual copies of guide markers for automatic content association.

Segmented sensor arrays track finger position and velocity across discrete elements to generate precise motion data.

Segmenting handwritten strokes into line segments and short arcs enables precise vector analysis of input data.

A neural network classifies handwritten payees on scanned checks by clustering non-white pixels and verifying against a secondary database.

A neural network model predicts personalized image sharing decisions using annotated images and user text input.

A simulated environment trains deep neural networks using virtual sensor data to control a virtual vehicle.

Fusing axial, coronal, and sagittal MRI data eliminates anisotropic resolution limits to reduce partial volume effects and improve boundary detection accuracy.

Retrieves similar images from a database to identify objects using associated tag information without prior training.

A classification system learns a linear projection to embed samples in a lower-dimensional space for nearest class mean analysis.

A multi-step image recognition framework classifies digital pathology images using sequential feature extraction.

Conductive wire labyrinths measure electrical properties to detect chassis tampering, ensuring examination security while maintaining test accessibility.

Portable computing devices define surveillance camera regions of interest simultaneously, eliminating separate computer systems for configuration.

Neural networks identify applied data augmentation types on sound waveforms to differentiate anomalous audio from normal samples despite dataset imbalance.

Manifold subspace segmentation trains specialized subclassifiers to detect anomalous input data points in machine learning anomaly detectors.

A contents providing system derives expendable item amounts based on printed image size to calculate precise printing costs.

A probable words dictionary integrates visual identification data to enhance voice recognition accuracy.

A work management apparatus tracks operator hand movements across segmented component areas to measure individual assembly process durations.

Segmenting findings by region count prioritizes single-region comments, reducing rewriting effort and accelerating interpretation report creation.

A fingerprint extraction method uses a guard region and predicted color distribution to isolate ridge patterns from camera-captured images.

Segmented pixel groups capture reflected light during distinct sub-periods within a single frame to calculate distance via phase differences.

A machine learning model extracts facial features from content items to generate deceptive scores for policy enforcement.

Accumulated watermarking pixels trace unauthorized video distribution without separate DRM files, resolving detection complexity.

Reference image comparison detects deviations from stored signal patterns, distinguishing static obstacles from moving objects and temporary blockages.

A vehicle display system processes camera images to maintain legible text orientation while mirroring the background environment.

Spectrum Mixup creates augmented data to resolve privacy conflicts while maintaining high face recognition accuracy across domains.

Mapping scanned document content onto a blank template corrects printing distortions and reduces storage space by eliminating redundant image data.

Logic program extraction assigns kernel labels by correlating activations with features.

A monitoring device determines detection time for second events based on first event timing in video data.

A visual analytics system processes video and mouse movement data to detect abnormal behavior during online exams.

A system filters environmental alerts by detecting objects and determining user location relative to those objects.

An image processing apparatus adjusts detection conditions based on face size to accelerate detection.

A camera processes image data through neural network layers on a first processing unit during sensor readout.

A compact neural architecture with residual modules accelerates deep learning model training through adaptive piecewise linear learning rate adjustments.

A calibration model corrects individual characteristics in user-assigned labels, resolving reliability issues caused by inconsistent manual data annotation.

Pruned artificial neural networks reduce computational resources while maintaining recognition accuracy.

A signal identification system uses convolutional neural networks and long short-term memory recurrent neural networks to process camera images.

Visible light image analysis detects smoke and fire by extracting focus distance, brightness, and color data without requiring expensive infrared filters.

A selection unit chooses compression parameters for composite images based on scanned and PDL data types.

A document processing system specifies disposition policies during reservation creation to streamline electronic conversion workflows.

Unsupervised deep neural network generates feature-specific bias vectors for automated model evaluation.

Information processing apparatus sets extraction rules using predetermined patterns, user-defined element arrangements, and coordinate inputs.

A fatigue degree calculating unit processes image and sound feature values to estimate user fatigue levels during video playback.

A magnetic garage door controller integrates wireless connectivity to automate operations based on user proximity.

Narrow colored lines within sinusoidal fringes resolve 2π ambiguity and interference, enabling precise unwrapped phase map computation.

Unsupervised machine learning clusters form components into structured representations using geometric attributes.