Top 30 Intermediate Computer Vision Concepts Quiz
30 intermediate multiple-choice Computer Vision concept questions, the mechanics underneath the basics: how the pieces relate and where the usual mental model stops holding. They come from 30 bites in the Computer Vision library, the middle slice of the 166 Computer Vision concept questions in the library. Answer them here or read straight down. Every question carries the correct option, why it is correct, and a link to the bite it came from.
Image/video models, diffusion, OCR, multimodal
30 questions. Pick an answer, or open “Show the answer” to read it.
Answers are graded in your browser. Nothing is saved, and no XP or streak is earned here. The app keeps score.
Question 1 of 30
Which camera property is NOT directly determined by camera resectioning?
Show the answer
Answer: a · The camera's internal lens distortion parameters.
Camera resectioning determines the camera's external pose (position and orientation) based on known 3D-to-2D point correspondences. It assumes internal properties like lens distortion are either known or handled by a separate calibration, making them not directly determined by resectioning. The 2D pixel locations of known 3D points are inputs to the process, not outputs.
Read the full bite: Camera Resectioning: Finding a Camera's Pose in 3D Space
Question 2 of 30
For which application is correcting lens distortion most essential?
Show the answer
Answer: a · Reconstructing a 3D environment from a series of images
Correcting lens distortion is mandatory for applications like 3D reconstruction (photogrammetry) that rely on precise geometric measurements from images. While object classification might use images, it often doesn't require the same geometric precision, and aesthetic quality can sometimes even be enhanced by distortion.
Read the full bite: Lens Distortion: Why Straight Lines Curve in Photos
Question 3 of 30
For which reason are HSL and HSV models generally unsuitable for image analysis and computer vision tasks?
Show the answer
Answer: b · They lack perceptual uniformity, making color difference measurements unreliable.
The card states that HSL and HSV are not perceptually uniform, meaning visual changes do not correspond consistently to numerical changes, which makes calculating color distance unreliable for algorithms. Other options are either incorrect or not the primary reason cited.
Read the full bite: HSL and HSV: Intuitive Ways to Represent RGB Color
Question 4 of 30
In an RGB system, what does "True color" (24-bit color depth) signify regarding bits per channel (bpc)?
Show the answer
Answer: d · Each of the Red, Green, and Blue channels uses 8 bits.
True color (24-bit color depth) means 24 bits per pixel (bpp). In an RGB system, these 24 bits are typically divided equally among the three channels (Red, Green, Blue), meaning 8 bits per channel (8 bpc). Option C is incorrect because 24 bits per channel would result in 72 bpp, not 24 bpp.
Read the full bite: Color Depth: Bits Per Pixel vs. Bits Per Channel
Question 5 of 30
A textile company needs to ensure fabric colors match a standard precisely, regardless of the display or printer used. Why would they primarily use CIELAB for this task?
Show the answer
Answer: b · It provides a device-independent measure of color perception, allowing for objective comparison and quality control.
CIELAB's primary advantage is its device-independent nature, modeling color based on human perception rather than display output. This allows for objective measurement and comparison of colors, crucial for quality control. Option C is less accurate because while CIELAB can be used for intuitive corrections, its core strength for precise matching lies in its perceptual uniformity and device independence, not just ease of adjustment.
Read the full bite: CIELAB Color Space: Measuring Color Beyond RGB
Question 6 of 30
In which scenario would applying a strong Gaussian blur be generally counterproductive or inappropriate?
Show the answer
Answer: a · When the goal is to preserve minute details and sharp boundaries in medical scans.
The card explicitly states that Gaussian blur should be avoided when preserving sharp edges and fine details, such as in medical imaging, is critical, as a strong blur destroys this high-frequency information. The other options (A, B, D) are all listed as appropriate and beneficial uses of Gaussian blur.
Read the full bite: Gaussian Blur: Smoothing Images with Weighted Averages
Question 7 of 30
When applying a median filter to an image, what is its key benefit compared to a simple averaging filter?
Show the answer
Answer: c · It effectively removes impulse noise while maintaining the sharpness of image edges.
The median filter's primary advantage is its ability to remove impulse noise (like salt-and-pepper noise) without blurring important image edges, which averaging filters tend to do. It is also a non-linear operation, not a linear one.
Read the full bite: Median Filter: Smoothing Images Without Blurring Edges
Question 8 of 30
Which of the following best describes the fundamental mechanism the Sobel operator uses to detect edges?
Show the answer
Answer: d · It calculates the gradient magnitude by measuring the rate of change in pixel brightness across small regions.
The Sobel operator functions as a discrete differentiation operator, measuring the local rate of change in pixel brightness (the gradient) to identify edges. Option B is a tempting distractor, but Sobel uses weighted differences via convolution kernels to approximate a derivative, not just simple direct comparisons.
Read the full bite: The Sobel Operator: Fast, Cheap Edge Detection
Question 9 of 30
To effectively remove both small bright specks and small dark holes from a binary image while preserving the main object's size, which operation sequence is most appropriate?
Show the answer
Answer: b · Opening followed by Closing
Opening (erosion then dilation) removes small bright specks (salt noise) without shrinking the main object. Subsequent Closing (dilation then erosion) fills small dark holes (pepper noise) without expanding the main object, making this sequence ideal for preserving overall object size.
Read the full bite: Morphological Transformations: Shaping Images with Kernels
Question 10 of 30
For which task would Scale-Space representation be most advantageous?
Show the answer
Answer: d · Detecting features in an image without prior knowledge of their specific scale or size.
Scale-space representation's primary purpose is to analyze image structures at all possible sizes, making it ideal for detecting features when their scale is unknown. Options B and D describe scenarios where the card explicitly states scale-space is unnecessary or computationally expensive, respectively. Option B describes a different image processing goal not directly addressed by scale-space's core benefit.
Read the full bite: Scale-Space: Analyzing Images at Multiple Scales
Question 11 of 30
The primary purpose of subtracting a heavily blurred image from a lightly blurred image in the Difference of Gaussians (DoG) technique is to:
Show the answer
Answer: a · Isolate features of a particular size by effectively removing both fine-grained noise and large, uniform regions.
The core idea of DoG is to isolate mid-sized details by canceling out both fine noise and large, uniform areas, as stated in the card's mental model. While DoG acts as a band-pass filter, it is an approximation and not intended for a precise mathematical representation of the frequency spectrum, making option D incorrect.
Read the full bite: Difference of Gaussians: Finding Edges by Subtracting Blurs
Question 12 of 30
For which scenario would FAST corner detection be most advantageous compared to other methods?
Show the answer
Answer: c · Tracking objects in a real-time video stream on a resource-limited device.
FAST is designed for speed, making it ideal for real-time applications on resource-constrained devices where low latency is critical. It is not robust to scale or rotation changes, nor is it prioritized for maximum accuracy in offline analysis.
Read the full bite: FAST: High-Speed Corner Detection for Real-Time Vision
Question 13 of 30
What is the core principle behind HOG's ability to describe object shape robustly against lighting variations?
Show the answer
Answer: b · It summarizes local gradient orientations and normalizes these summaries within overlapping blocks.
The card explains that HOG computes gradient directions to capture an object's shape and normalizes these summaries within overlapping blocks to achieve robustness against lighting changes. The other options describe mechanisms not central to HOG's operation, such as color analysis, average intensity, or texture patterns.
Question 14 of 30
What is the most significant advantage of using SIFT features for image matching tasks?
Show the answer
Answer: c · They provide descriptors that are robust to changes in image scale.
SIFT's primary strength, as highlighted in the card, is its ability to find and describe features that remain stable despite changes in image scale. The card explicitly warns against the misconception that SIFT is invariant to all changes, making option B a tempting but incorrect distractor.
Question 15 of 30
Which scenario would most likely lead to inaccurate results if only a single homography is used?
Show the answer
Answer: a · Stitching together multiple images to create a panorama of a distant mountain range.
A homography is designed to map points between images of a flat surface. A distant mountain range, despite appearing far, still presents significant depth and parallax, which a single homography cannot accurately model, as explicitly stated in the 'When Not To Use It' section. The other options describe scenarios involving flat surfaces where a homography is appropriate.
Read the full bite: Homography: Mapping Flat Surfaces Between Images
Question 16 of 30
What is the primary benefit of employing the Fundamental Matrix in computer vision tasks involving two images?
Show the answer
Answer: a · It reduces the search space for a corresponding point in the second image from a 2D area to a 1D line.
The Fundamental Matrix's core function is to constrain the search for a corresponding point in a second image to an epipolar line, thereby reducing a 2D search to a more efficient 1D search. Option C describes the Essential Matrix, which is used for calibrated cameras and provides metric information, unlike the Fundamental Matrix which works with uncalibrated cameras and projective geometry.
Read the full bite: The Fundamental Matrix: Constraining 3D Search to a Line
Question 17 of 30
What is a crucial prerequisite for effectively using the Essential Matrix in 3D vision tasks?
Show the answer
Answer: b · The intrinsic parameters of the cameras must be known.
The card explicitly states that the Essential Matrix should be used "when you know the intrinsic parameters of your camera." Using it with uncalibrated cameras is identified as "the biggest mistake," making option A incorrect.
Read the full bite: The Essential Matrix: Geometry for Stereo Vision
Question 18 of 30
What is the most critical prerequisite for successfully performing triangulation to determine a 3D point?
Show the answer
Answer: b · Precise knowledge of each camera's internal and external parameters.
The card states that knowing the 3D position, orientation, and lens properties of each camera (its 'camera matrix') is essential to trace rays and find the 3D point. While accurate feature matching is crucial for good results, the method accounts for noise, meaning perfect matching isn't a strict prerequisite for the algorithm to function.
Read the full bite: Triangulation: Finding 3D Points from 2D Images
Question 19 of 30
Which scenario would present the greatest challenge for Structure from Motion (SfM) to accurately reconstruct a 3D model?
Show the answer
Answer: b · A modern building facade made entirely of large, uniform glass panels
SfM struggles with scenes lacking texture or featuring highly reflective/transparent surfaces, as it relies on detecting and matching distinct keypoints. Uniform glass panels fit this description perfectly, making it difficult for SfM to find stable points for triangulation, unlike the textured surfaces in the other options.
Read the full bite: Structure from Motion (SfM): Rebuilding 3D from 2D Photos
Question 20 of 30
What is the primary benefit of ReLU over sigmoid and tanh, particularly in deep neural networks?
Show the answer
Answer: c · It prevents gradients from diminishing to zero for positive inputs during backpropagation.
The card explicitly states that older functions suffered from the "vanishing gradient" problem and that ReLU's constant '1' derivative for positive values "prevents the gradient from vanishing". While ReLU is faster (option B), the primary problem it solves for deep networks, as explained in the "WHY IT EXISTS" section, is the vanishing gradient problem, which directly hinders learning.
Read the full bite: ReLU: The 'On/Off' Switch for Neural Networks
Question 21 of 30
What is the primary reason cross-entropy loss is preferred over simple accuracy for training classification models?
Show the answer
Answer: c · It offers a continuous measure of error that heavily penalizes confident, incorrect predictions.
The card states that cross-entropy provides a continuous measure of error that heavily penalizes confident, wrong predictions, which is crucial for guiding model training effectively. Simple accuracy (Option D) only indicates right or wrong, lacking this nuanced error signal.
Read the full bite: Cross-Entropy Loss: How Wrong Is Your Model's Guess?
Question 22 of 30
What mathematical principle is fundamental to backpropagation's ability to efficiently adjust neural network weights?
Show the answer
Answer: c · The chain rule of calculus for gradient computation.
The card states that backpropagation "Using the chain rule from calculus, it calculates the gradient of the loss with respect to every single weight in the network." This mathematical principle is crucial for efficiently determining how each weight contributes to the overall error. While other mathematical concepts are used in neural networks, the chain rule is central to the backward pass of backpropagation. Matrix inversion is not used for weight updates in gradient-based optimization.
Read the full bite: Backpropagation: How Neural Networks Learn from Mistakes
Question 23 of 30
What is the main objective of applying data augmentation techniques in deep learning for computer vision?
Show the answer
Answer: d · To improve the model's ability to generalize to new, unseen data by reducing overfitting.
Data augmentation's primary goal is to combat overfitting, especially with small datasets, by creating varied versions of existing images. This forces the model to learn general features rather than memorizing specific training examples, thereby improving its performance on new data. Option A describes the problem of overfitting, not the solution.
Read the full bite: Data Augmentation: Getting More Images for Free
Question 24 of 30
How does Dropout primarily enable a neural network to generalize better?
Show the answer
Answer: a · By forcing individual neurons to learn more robust features independently.
The card explains that Dropout "forces the network to learn more robust features instead of relying on specific neurons" and uses the analogy of "forcing every expert to become more capable on their own." This mechanism of encouraging independent feature learning is how it primarily improves generalization. Option D describes a common interpretation of Dropout's effect, but the direct mechanism is the forced independence of neurons, not the explicit creation of multiple voting networks.
Read the full bite: Dropout: Forcing a Network to Generalize
Question 25 of 30
What was the primary factor enabling AlexNet's breakthrough in image recognition?
Show the answer
Answer: c · The synergistic combination of deep convolutions with ReLU, dropout, and GPU training.
AlexNet's success was not due to a single innovation but rather the effective combination of several techniques, including deep convolutional layers, ReLU activation, dropout, and GPU parallelization. The card explicitly states its novelty was 'combining new techniques' and describes it as a 'brilliant recipe' of ingredients.
Read the full bite: AlexNet: The CNN That Sparked the Deep Learning Boom
Question 26 of 30
What primary bottleneck in early object detection did the Region Proposal Network (RPN) aim to solve?
Show the answer
Answer: b · The slow and separate process of generating potential object regions.
The card explicitly states that the 'initial region proposal step became the major performance bottleneck' and the RPN was created to solve this by integrating proposal generation directly. Option C describes the classification stage, which is a subsequent step, not the RPN's primary concern.
Read the full bite: Region Proposal Network (RPN): The 'Where to Look' Engine
Question 27 of 30
What is the primary architectural innovation that enables YOLO's real-time object detection capability?
Show the answer
Answer: a · It uses a single convolutional neural network to process the entire image in one forward pass.
The card states YOLO's innovation is treating detection as a "single regression problem" by processing the image "once" through a "single convolutional neural network" in "one forward pass." This contrasts with previous two-stage systems, making option C incorrect as it describes the older, slower approach.
Read the full bite: YOLO: Real-Time Object Detection in a Single Pass
Question 28 of 30
How do Fully Convolutional Networks (FCNs) primarily achieve pixel-level labeling for semantic segmentation?
Show the answer
Answer: b · By replacing fully-connected layers with convolutions and employing upsampling with skip connections.
FCNs achieve pixel-level labeling by replacing traditional fully-connected layers with convolutions to preserve spatial data and then using upsampling layers, often with skip connections, to restore the original image dimensions. Option A describes the inefficient method FCNs were designed to replace.
Read the full bite: Fully Convolutional Networks: From Image to Segmentation Map
Question 29 of 30
For which application is U-Net the most appropriate architecture?
Show the answer
Answer: c · Tracing precise cell boundaries in 512x512 microscopy images with only 200 training masks
U-Net is built for pixel-accurate segmentation when training data is limited, exactly as described in option C. Option B is simple classification, B is bounding-box detection on high-resolution footage, and D combines massive data with only coarse localization needs, all scenarios the card identifies as poor fits for U-Net.
Question 30 of 30
What is the primary advantage of using dilated convolutions over traditional pooling layers in tasks like semantic segmentation?
Show the answer
Answer: a · It allows the network to process a wider context while maintaining high spatial resolution.
Dilated convolution expands the receptive field by skipping pixels, allowing the model to see a broader context without downsampling, which preserves the high spatial resolution crucial for pixel-level tasks. Option B is incorrect because dilated convolution explicitly skips pixels, leading to less dense local sampling, not denser.
Read the full bite: Dilated Convolution: A Wider View Without More Parameters
Could you explain these out loud?
That is what an interview actually tests. Tezvyn gives you questions like these with what the interviewer is really checking, the answer that lands, and the mistake that ends the conversation, in the four minutes before your next meeting.
The iPhone app is on the way
We are building it. Until it lands, nothing here is held back from you: every interview card, your saved cards, streaks and the job board all work in Safari, plus hundreds of free practice quizzes of thirty questions each. Sign in and it all carries over to the app the day it arrives.
Want it as an icon? Tap Share at the bottom of Safari, then Add to Home Screen. It opens full screen and the cards you have read stay available offline.