Skip to main content
#technical-english#computer-vision#vocabulary#quiz

Essential Vocabulary

Study computer vision and machine learning terms with English definitions, worked examples, and a ten-question vocabulary quiz.

#Core Concepts and Terminology

#1. Perception and Processing

  • Perceive (verb): To notice or interpret something through the senses.

    • Example: Computers can perceive subtle differences in images that humans might miss.
  • Perceptual (adjective): Relating to how sensory information is interpreted.

    • Example: Perceptual computing aims to mimic human sensory processing.
  • Misperception (noun): An incorrect impression or interpretation.

    • Example: Image artifacts can lead to misperception in computer vision systems.

#2. Visual Properties

  • Translucency (noun): The property of letting light through while scattering it so details appear blurred.

    • Example: Processing translucency in images requires advanced rendering techniques.
  • Subtle (adjective): Slight or delicate enough to be hard to notice.

    • Example: The algorithm can detect subtle changes in lighting conditions.
  • Shading (noun): Variations in brightness that show how light falls on a surface.

    • Example: Proper shading analysis helps in understanding 3D structures.

#3. Image Analysis

  • Effortlessly (adverb): With little difficulty or exertion.

    • Example: Modern AI systems effortlessly process thousands of images per second.
  • Portrait (noun): A photograph or other likeness of a person.

    • Example: Face detection algorithms are crucial for portrait photography.
  • Rapid (adjective): Moving or happening quickly.

    • Example: Rapid image processing is essential for real-time applications.

#4. Technical Processes

  • Delineate (verb): To trace an outline or describe something precisely.

    • Example: The algorithm can delineate object boundaries precisely.
  • Causality (noun): A relationship in which one event brings about another.

    • Example: Understanding causality in scene analysis improves prediction accuracy.
  • Radiometry (noun): The measurement of electromagnetic radiation, including visible light.

    • Example: Radiometry is fundamental to understanding light behavior in images.

#5. Industrial Applications

  • Inspection (noun): A close examination to check condition or quality.

    • Example: Automated visual inspection systems in manufacturing.
  • Retail (noun): The sale of goods directly to consumers.

    • Example: Computer vision enhances retail analytics and customer behavior tracking.
  • Warehouse (noun): A building where goods are stored.

    • Example: Automated warehouse systems use computer vision for inventory management.
  • Logistics (noun): The planning and movement of goods through a supply chain.

    • Example: Computer vision optimizes logistics operations through package tracking.

#6. Advanced Applications

  • Intra-operative (adjective): Taking place during surgery.

    • Example: Intra-operative imaging assists surgeons during procedures.
  • Morphology (noun): The study of form and structure.

    • Example: Mathematical morphology is used in image processing.
  • Surveillance (noun): Monitoring people, objects, or places over time.

    • Example: Video surveillance systems employ advanced computer vision techniques.

#7. Image Processing Techniques

  • Stitching (noun): Joining several images to make a single image.

    • Example: Panorama creation through image stitching.
  • Overlapping (adjective): Covering some of the same area.

    • Example: Overlapping images are used in 3D reconstruction.
  • Bracketing (noun): Taking several exposures of the same scene at different settings.

    • Example: HDR images are created using exposure bracketing.
  • Morphing (noun): Gradually transforming one image or shape into another.

    • Example: Face morphing creates smooth transitions between images.

#8. Research and Development

  • Meticulously (adverb): With careful attention to detail.

    • Example: AI models must be meticulously trained for optimal performance.
  • Layer-wise, sample-wise (adjective): Applied separately to each layer or to each sample.

    • Example: Neural networks are analyzed layer-wise for better understanding.
  • Preexisting (adjective): Already present before a particular event or step.

    • Example: Preexisting models can be fine-tuned for specific tasks.
  • Plausible (adjective): Reasonable or believable in light of the evidence.

    • Example: The system generates plausible reconstructions of 3D scenes.

#9. Industrial Human-Robot Collaboration

  • Pose Forecasting (noun phrase): Predicting how the position and orientation of a body or its joints will change.
    • Example: Pose forecasting is crucial for safe human-robot interaction.
  • Ablation (noun): Removing a component of a model to test its contribution.
    • Example: Ablation studies help understand the contribution of different model components.
  • Endow (verb): To give someone or something a particular capability.
    • Example: The new sensor endows the robot with better perceptive capabilities.
  • Perceptive (adjective): Quick to notice or understand something.
    • Example: Perceptive robots can adapt to dynamic environments.
  • Harm (noun or verb): Damage or injury, or the act of causing it.
    • Example: The primary goal is to prevent any harm to the human worker.
  • Transmission of contact wrenches (noun phrase): The transfer of forces and moments through a point or surface of contact.
    • Example: Accurate transmission of contact wrenches is vital for haptic feedback.
  • On-the-fly (adverbial phrase): During operation, as the need arises.
    • Example: The robot can adjust its path on-the-fly to avoid obstacles.
  • Adjacency (noun): The relationship between neighboring or directly connected elements.
    • Example: Adjacency matrices are used to represent connections in a graph.
  • Sparse (adjective): Containing few occupied positions or few nonzero entries.
    • Example: Sparse data can be challenging for some machine learning models.
  • Sparsity (noun): The property of having relatively few nonzero or occupied elements.
    • Example: Sparsity is a desirable property in many high-dimensional datasets.
  • Abide (verb): To follow a rule or meet a requirement, as in “abide by.”
    • Example: Robots must abide by safety protocols.
  • Genuine (adjective): Authentic or real.
    • Example: The system aims to achieve genuine collaboration between humans and robots.
  • Foresee (verb): To anticipate something before it occurs.
    • Example: It is difficult to foresee all possible scenarios in a complex environment.
  • Exploited (verb (past tense or past participle)): Made use of for a particular purpose.
    • Example: The robot’s capabilities were fully exploited in the task.
  • Aspect (noun): One side or feature of a subject.
    • Example: Safety is a critical aspect of human-robot collaboration.
  • Intersection (noun): A point, region, or set of elements shared by two or more things.
    • Example: The intersection of AI and robotics has led to many innovations.
  • Prune (verb): To trim or remove unnecessary parts.
    • Example: Pruning the neural network can reduce its complexity.
  • Factorize (verb): To express a quantity or matrix as a product of factors.
    • Example: We can factorize the matrix to understand its underlying structure.
  • Adjacency (noun): The relationship between neighboring or directly connected elements. This term also appears earlier in the list.
    • Example: Adjacency lists are another way to represent graph connections.
  • Spectral (adjective): Relating to a spectrum, such as a range of wavelengths or eigenvalues.
    • Example: Spectral analysis is used in various sensor technologies.
  • Coarse (adjective): Approximate or low in detail or resolution.
    • Example: A coarse estimation was made initially, followed by a finer adjustment.
  • Sparsifying (verb or adjective): Making a representation contain fewer nonzero entries.
    • Example: Sparsifying techniques are used to reduce data dimensionality.

#10. Modern Artificial Intelligence

  • Formalize (verb): To express an idea in a precise mathematical or logical form.
    • Example: We need to formalize the problem statement before designing a solution.
  • Density (noun): The amount of a quantity per unit of space or another measure.
    • Example: Probability density functions describe the likelihood of a continuous variable.
  • PDF (Probability Density Function) (noun): A function whose integral over a range gives the probability of a continuous variable falling there.
    • Example: The PDF of a normal distribution is bell-shaped.
  • PMF (Probability Mass Function) (noun): A function that assigns a probability to each possible value of a discrete random variable.
    • Example: The PMF is used for discrete random variables.
  • Accomplish (verb): To complete a task or achieve a goal.
    • Example: The AI model was able to accomplish the task with high accuracy.
  • Observed (adjective or past-tense verb): Seen, measured, or recorded.
    • Example: The observed data was used to train the model.
  • Intractable (adjective): Difficult to handle or computationally impractical to solve.
    • Example: Some problems are computationally intractable.
  • Coefficient (noun): A factor that multiplies a variable or term.
    • Example: The coefficients of the linear regression model were estimated.
  • Era (noun): A distinct period of time.
    • Example: We are in the era of big data and artificial intelligence.
  • Deluge (noun): A flood or an overwhelming influx.
    • Example: Researchers face a deluge of data from various sources.
  • Apparently (adverb): Seemingly, or as the available evidence suggests.
    • Example: Apparently, the new algorithm outperforms the old one.
  • Sufficient (adjective): Enough to meet the stated need.
    • Example: Sufficient data is required to train a robust model.
  • Altering (verb or adjective): Changing or modifying something.
    • Example: Altering the hyperparameters can significantly affect model performance.
  • Ability to fix (phrase): The capacity to correct a fault or solve a problem.
    • Example: The system has the ability to fix minor errors automatically.
  • Devising (verb): Developing or designing a method or solution.
    • Example: Devising new algorithms is a key part of AI research.
  • Empirical (adjective): Based on observation or experiment.
    • Example: Empirical results show the effectiveness of the proposed method.
  • Theorem (noun): A mathematical or logical statement established by proof.
    • Example: The Bayes’ theorem is fundamental in probability theory.
  • Criterion (noun): A standard for making a judgment. Its plural is “criteria.”
    • Example: Accuracy is a common criterion for evaluating classification models.
  • Impose (verb): To apply or enforce a condition, rule, or constraint.
    • Example: We can impose constraints on the optimization problem.
  • Regularization (noun): A technique that penalizes or restricts a model to reduce overfitting.
    • Example: Regularization helps prevent overfitting in machine learning models.
  • Implicitly (adverb): In a way that is implied or learned without being stated directly.
    • Example: The model implicitly learns features from the data.
  • Explicitly (adverb): In a way that is stated directly and clearly.
    • Example: Some parameters need to be explicitly defined.
  • Preferences (noun): Choices or outcomes that someone favors over others.
    • Example: The recommendation system learns user preferences over time.
  • Validation (noun): Checking whether a result or a model meets the intended requirements.
    • Example: Cross-validation is used to assess model performance.
  • Problematic (adjective): Likely to cause difficulties or raise concerns.
    • Example: Handling missing data can be problematic.
  • Associate (verb): To connect one thing with another.
    • Example: The goal is to associate input features with output labels.
  • Evaluations (noun): Assessments of quality, performance, or effectiveness.
    • Example: Multiple evaluations were conducted to ensure robustness.
  • Converge (verb): To approach a common point or a limiting value.
    • Example: The optimization algorithm should converge to a good solution.
  • Convex (adjective): Describing a set containing the segment between any two of its points, or a function whose graph lies below its chords.
    • Example: Convex optimization problems are generally easier to solve.

#Writing Technical Papers

  • Inventing (verb): Creating or developing something new.

    • Example: Inventing new algorithms for efficient image processing.
  • Revising (verb): Editing or reworking an existing text or design.

    • Example: Revising the methodology section of the research paper.
  • Drafting (verb): Writing an initial version.

    • Example: Drafting the experimental results section.

#11. Mathematics, Linear Algebra, Probability, and Machine Learning

  • Vector: A mathematical object with magnitude and direction, commonly represented by an ordered list of numbers.
  • Matrix: A rectangular array of numbers arranged in rows and columns.
  • Inner Product (Dot Product): An operation on two vectors that produces a scalar and can measure their alignment or similarity.
  • Outer Product: An operation that multiplies a column vector by a row vector to produce a matrix.
  • Matrix-Vector Product: Multiplying a matrix by a vector produces a vector, a linear combination of the columns of the matrix.
  • Matrix-Matrix Product: Multiplying two matrices of compatible dimensions to obtain another matrix.
  • Identity Matrix: A square matrix with ones on its main diagonal and zeros elsewhere. It acts as the multiplicative identity.
  • Diagonal Matrix: A square matrix whose entries off the main diagonal are zero.
  • Transpose: The matrix obtained by exchanging rows and columns, written Aᵀ.
  • Symmetric Matrix: A square matrix equal to its transpose, A = Aᵀ.
  • Trace: The sum of the entries on the main diagonal of a square matrix, written tr(A).
  • Norm: A function that measures the size of a vector or matrix. Examples include the L1, L2, infinity, and Frobenius norms.
  • Linear Independence: The property that no vector in a collection can be written as a linear combination of the others.
  • Rank: The maximum number of linearly independent rows or columns of a matrix. It gives the dimension of its image.
  • Invertible (Non-singular) Matrix: A square matrix for which an inverse exists.
  • Inverse Matrix: The unique matrix A⁻¹ that gives the identity when multiplied by A on either side.
  • Orthogonal Vectors: Two vectors whose inner product is zero. In Euclidean space their directions are perpendicular.
  • Normalized Vector: A vector scaled to have an L2 norm of one.
  • Orthogonal Matrix: A real square matrix U with orthonormal columns, so UᵀU = UUᵀ = I and its inverse equals its transpose.
  • Determinant: A scalar for a square matrix whose magnitude gives the volume scale factor of its linear transformation. It is zero for a singular matrix.
  • Quadratic Form: A scalar expression of the form xᵀAx, where A is square and x is a vector.
  • Positive Definite (PD) Matrix: A real symmetric matrix A for which xᵀAx > 0 for every nonzero vector x.
  • Positive Semidefinite (PSD) Matrix: A real symmetric matrix A for which xᵀAx ≥ 0 for every vector x.
  • Eigenvalue: A scalar λ for which Ax = λx for some nonzero vector x. Multiplying by A scales that vector by λ.
  • Eigenvector: A nonzero vector x satisfying Ax = λx for its associated eigenvalue λ.
  • Gradient: The vector of first partial derivatives, ∇f(x). It points in the direction of steepest increase when it is nonzero.
  • Hessian: The matrix of second partial derivatives, ∇²f(x). It describes local curvature and helps assess convexity or concavity.
  • Random Variable: A function that assigns a numerical value to each outcome of an experiment.
  • Cumulative Distribution Function (CDF): The probability that X is at most x, written F_X(x) = P(X ≤ x).
  • Probability Density Function (PDF): A function describing an absolutely continuous distribution. Where the derivative exists, p_X(x) = dF_X(x)/dx.
  • Independent Random Variables: Variables whose joint distribution factors into the product of their marginal distributions.
  • Expected Value (Mean): The average under the probability distribution. For a continuous variable, E[X] = ∫ x p_X(x) dx when the expectation exists.
  • Variance: A measure of dispersion around the mean, Var(X) = E[|X − E[X]|²].
  • Correlation: A measure of association between two random variables. In the signal-processing convention used here, the uncentered correlation is E[XY*], where * denotes complex conjugation.
  • Covariance: A measure of how two variables vary together, Cov(X,Y) = E[(X − E[X])(Y − E[Y])*].
  • Random Vector: A vector whose entries are random variables.
  • Multivariate Gaussian (Normal) Distribution: A joint normal distribution for several variables, specified by a mean vector and a covariance matrix.
  • Jointly Gaussian Random Vector: A random vector for which every linear combination of its entries is Gaussian. Gaussian marginal distributions alone do not guarantee this property.
  • Random Process: A family of random variables indexed by time or space. This guide considers processes on a two-dimensional discrete spatial grid.
  • Power Spectrum (Power Spectral Density): The discrete-space Fourier transform of the autocorrelation of a wide-sense stationary random process. It describes how power is distributed across spatial frequencies.
  • Loss Function: A function L(D(x), y) that measures the discrepancy between a model prediction D(x) and a target y.
  • Supervised Learning: Training a model on examples that pair inputs x with known targets y.
  • Empirical Loss Function: The average or sum of losses on observed training samples, used to approximate loss under the data distribution.
  • Gradient Descent: An optimization method that moves in the direction of the negative gradient to seek a local minimum.
  • Learning Rate (Step Size): The factor that scales the gradient at each update and therefore controls the size of the step.
  • Stochastic Gradient Descent (SGD): An optimization method that updates parameters using a gradient estimated from a randomly sampled example or mini-batch.
  • Batch Size: The number of examples used for each parameter update in mini-batch training.
  • Local Minimum: A point whose function value is no greater than the values at nearby points. A zero gradient and a positive definite Hessian give a sufficient condition for a strict local minimum.
  • Overfitting: Fitting the training data so closely that performance on new data suffers.
  • Underfitting: Failing to capture the underlying pattern, resulting in poor performance on both training and test data.
  • Regularization: A modification to the learning procedure that discourages overfitting and can improve generalization.
  • No Free Lunch Theorem: A result showing that no single learning algorithm performs best across all possible problems without assumptions about the problems.
  • Cross-Validation: Estimating generalization by splitting data into folds and rotating which folds are used for training and evaluation.
  • Classification: Assigning an input to one of a predefined set of classes.
  • Regression: Predicting a continuous numerical output from input data.
  • Density Estimation: Modeling or estimating the probability distribution of observed data.
  • Object Detection: Locating objects in an image or video and identifying their classes, often combining localization by regression with classification.
  • Self-Supervised Learning: Training on a signal constructed from the input data itself, without manually supplied target labels.

Connect