Site icon T4Tutorials.com

Semantic & Instance Segmentation (Mask R-CNN, U-Net) — MCQs | Digital Image Processing

1. What is the primary goal of semantic segmentation in image processing?

(A) Detecting edges in images


(B) Assigning a class label to each pixel


(C) Enhancing image contrast


(D) Converting color images to grayscale



2. Which segmentation technique assigns different labels to distinct object instances of the same class?

(A) Semantic segmentation


(B) Edge detection


(C) Instance segmentation


(D) Thresholding



3. Which deep learning model is widely used for instance segmentation?

(A) ResNet


(B) Mask R-CNN


(C) VGGNet


(D) AlexNet



4. Which architecture is commonly used for biomedical image segmentation tasks?

(A) U-Net


(B) R-CNN


(C) YOLO


(D) DenseNet



5. In Mask R-CNN, what component is responsible for predicting object masks?

(A) Feature Pyramid Network


(B) Region Proposal Network


(C) ROIAlign


(D) Mask branch



6. What does U-Net use to combine low-level and high-level features?

(A) Dense connections


(B) Skip connections


(C) Pooling layers


(D) Residual blocks



7. Which part of U-Net architecture is responsible for upsampling the feature maps?

(A) Contracting path


(B) Bottleneck


(C) Expanding path


(D) Residual path



8. What is ROIAlign used for in Mask R-CNN?

(A) Generating bounding boxes


(B) Detecting object classes


(C) Accurate spatial alignment


(D) Downsampling feature maps



9. Which loss function is commonly used for pixel-wise classification in segmentation?

(A) Cross-entropy loss


(B) Triplet loss


(C) Mean squared error


(D) Contrastive loss



10. What is a major limitation of semantic segmentation?

(A) Cannot detect object boundaries


(B) Cannot handle multiple classes


(C) Cannot distinguish between instances of the same class


(D) Cannot process RGB images



11. What advantage does instance segmentation provide over semantic segmentation?

(A) Faster training


(B) Fewer parameters


(C) Better object counting and tracking


(D) Easier implementation



12. Which layer in U-Net is responsible for downsampling the input image?

(A) Transpose convolution


(B) Max pooling


(C) Batch normalization


(D) Softmax



13. In semantic segmentation, which class is often used to represent background pixels?

(A) Class 1


(B) Class 0


(C) Class 255


(D) Class -1



14. What is the output size of a segmentation model for an image of size 256×256?

(A) 1×1


(B) 256×256


(C) 64×64


(D) 512×512



15. Why is skip connection used in U-Net?

(A) To reduce training time


(B) To improve gradient flow


(C) To combine low-level spatial features with high-level semantic features


(D) To regularize the model



16. Which component in Mask R-CNN is modified from Faster R-CNN to enable segmentation?

(A) Backbone network


(B) Region proposal network


(C) Classifier head


(D) Additional mask prediction head



17. What type of data is required to train a semantic segmentation model?

(A) Image-level labels


(B) Pixel-level annotations


(C) Bounding boxes


(D) Object counts



18. What does the term “end-to-end” mean in the context of segmentation models?

(A) Model runs only on CPUs


(B) Training is done in multiple stages


(C) Full pipeline is trained in a single unified model


(D) Only inference is performed



19. Which evaluation metric is commonly used for segmentation tasks?

(A) Accuracy


(B) Mean Average Precision


(C) Intersection over Union (IoU)


(D) F1 Score



20. What is the role of data augmentation in training segmentation models?

(A) Reduces model accuracy


(B) Improves generalization


(C) Slows down training


(D) Reduces dataset size



21. What kind of masks does Mask R-CNN generate for detected objects?

(A) Binary masks


(B) Grayscale masks


(C) RGB masks


(D) Alpha masks



22. Which model extends Faster R-CNN for segmentation tasks?

(A) YOLOv5


(B) U-Net++


(C) Mask R-CNN


(D) PSPNet



23. What does the “contracting path” in U-Net primarily consist of?

(A) Upsampling layers


(B) Convolution and pooling layers


(C) Recurrent units


(D) Residual connections



24. What kind of convolution is used to upsample feature maps in U-Net?

(A) Depthwise convolution


(B) Standard convolution


(C) Transposed convolution


(D) Dilated convolution



25. Which activation function is typically used at the final layer of a binary segmentation model?

(A) ReLU


(B) Sigmoid


(C) Tanh


(D) Softmax



26. Which activation function is used at the final layer of a multi-class semantic segmentation model?

(A) Sigmoid


(B) Softmax


(C) ReLU


(D) Tanh



27. What problem does semantic segmentation aim to solve in image processing?

(A) Object detection


(B) Classifying each pixel in the image


(C) Enhancing image resolution


(D) Removing noise



28. In segmentation, what is the meaning of class imbalance?

(A) Equal number of samples per class


(B) More classes than objects


(C) Some classes have fewer labeled pixels


(D) Same mask for all objects



29. Which of the following models is NOT primarily used for segmentation?

(A) U-Net


(B) SegNet


(C) Mask R-CNN


(D) ResNet



30. Which backbone is commonly used in Mask R-CNN?

(A) MobileNet


(B) ResNet


(C) VGG16


(D) Inception



31. How does Mask R-CNN improve spatial alignment compared to Faster R-CNN?

(A) Uses ROIAlign instead of ROIPool


(B) Removes max pooling


(C) Uses feature pyramids


(D) Uses depth-wise convolutions



32. Which model uses a U-shaped architecture for segmentation?

(A) YOLOv4


(B) DeepLab


(C) U-Net


(D) Mask R-CNN



33. What is one limitation of U-Net?

(A) Cannot handle grayscale images


(B) Not suitable for large images without tiling


(C) Requires more than 10 GPUs


(D) Does not support skip connections



34. In segmentation, what is meant by a “mask”?

(A) Image caption


(B) Region of interest


(C) Binary or labeled pixel map


(D) Feature vector



35. Which layer increases the spatial resolution of features in U-Net?

(A) Max pooling


(B) Dropout


(C) Up-convolution


(D) Flatten



36. Which problem arises due to overlapping objects in instance segmentation?

(A) Poor bounding boxes


(B) Low image resolution


(C) Ambiguous mask assignment


(D) Gradient vanishing



37. How is multi-class segmentation different from binary segmentation?

(A) Operates only on grayscale images


(B) Uses multiple input images


(C) Predicts more than two classes per pixel


(D) Requires bounding boxes



38. What does a pixel-wise classification involve?

(A) Assigning RGB values to each pixel


(B) Predicting class labels for individual pixels


(C) Estimating camera parameters


(D) Calculating object velocity



39. Which of the following is a challenge in semantic segmentation?

(A) Object tracking


(B) Class overlap


(C) Small object detection


(D) Reducing color channels



40. What does “fine-grained segmentation” refer to?

(A) Classifying only large objects


(B) Coarse labeling of regions


(C) Precise boundaries and detailed labeling


(D) Only segmenting backgrounds



41. What is the function of the decoder in U-Net?

(A) Extract low-level features


(B) Reduce spatial dimensions


(C) Reconstruct segmentation map


(D) Normalize data



42. Which image modality is U-Net particularly effective for?

(A) Natural scene images


(B) Aerial images


(C) Medical images


(D) Infrared images



43. What is the role of batch normalization in segmentation networks?

(A) Removes noise from input


(B) Stabilizes and speeds up training


(C) Compresses the model


(D) Downsamples the image



44. Which type of data labeling is required for training Mask R-CNN?

(A) Image-level labels only


(B) Bounding boxes only


(C) Pixel-wise masks and bounding boxes


(D) Captions for each object



45. Which model applies atrous (dilated) convolution for segmentation tasks?

(A) YOLOv3


(B) DeepLab


(C) Mask R-CNN


(D) AlexNet



46. Why is semantic segmentation important in autonomous driving?

(A) To reduce latency


(B) To compress video data


(C) To identify road areas and obstacles


(D) To track object motion



47. What is one benefit of using data augmentation in segmentation tasks?

(A) Lower memory usage


(B) Faster inference


(C) Improved model generalization


(D) Increased overfitting



48. What is an important property of the masks predicted by Mask R-CNN?

(A) Grayscale masks for better depth


(B) Pixel-wise alignment with detected objects


(C) Class labels for the whole image


(D) Fixed size regardless of object scale



49. What is the main purpose of the Region Proposal Network (RPN) in Mask R-CNN?

(A) Segment objects pixel-by-pixel


(B) Generate feature maps


(C) Propose candidate object bounding boxes


(D) Upsample the image resolution



50. Which challenge is commonly addressed using post-processing in segmentation models?

(A) Reducing number of classes


(B) Aligning image histogram


(C) Refining mask boundaries and removing noise


(D) Increasing training loss



More MCQs on Digital image Processing

  1. Introduction to DIP — MCQs | Digital Image Processing

  2. Human Visual System (HVS) — MCQs | Digital Image Processing

  3. Image Acquisition Devices — MCQs | Digital Image Processing

  4. Image Sampling & Quantization — MCQs | Digital Image Processing

  5. Image Resolution & Bit Depth — MCQs | Digital Image Processing

  6. Basic Image Operations (Negative, Log, Power-law) — MCQs | Digital Image Processing

  7. Histogram Equalization & Specification — MCQs | Digital Image Processing

  8. Contrast Stretching — MCQs | Digital Image Processing

  9. Image Arithmetic (Add, Subtract, Multiply, Divide) — MCQs | Digital Image Processing

  10. Bit-plane Slicing — MCQs | Digital Image Processing

  11. Smoothing Filters (Mean, Gaussian, Median) — MCQs | Digital Image Processing

  12. Sharpening Filters (Laplacian, Gradient) — MCQs | Digital Image Processing

  13. High-Boost Filtering — MCQs | Digital Image Processing

  14. Edge Detection (Sobel, Prewitt, Roberts, Canny, LoG) — MCQs | Digital Image Processing

  15. Fourier Transform (DFT, FFT) — MCQs | Digital Image Processing

  16. Frequency Domain Filtering — MCQs | Digital Image Processing

  17. Low-pass & High-pass Filters — MCQs | Digital Image Processing

  18. Homomorphic Filtering — MCQs | Digital Image Processing

  19. Noise Models (Gaussian, Salt & Pepper, Speckle) — MCQs | Digital Image Processing

  20. Adaptive Filtering — MCQs | Digital Image Processing

  21. Inverse & Wiener Filtering — MCQs | Digital Image Processing

  22. Pseudo-color & True-color Processing — MCQs | Digital Image Processing

  23. Color Space Conversion (RGB ↔ HSV, HSI, YCbCr) — MCQs | Digital Image Processing

  24. Color Image Enhancement — MCQs | Digital Image Processing

  25. Image Segmentation (Thresholding, Otsu, K-means, Region Growing) — MCQs | Digital Image Processing

  26. Edge-based Segmentation — MCQs | Digital Image Processing

  27. Region Splitting and Merging — MCQs | Digital Image Processing

  28. Watershed Algorithm — MCQs | Digital Image Processing

  29. Morphological Operations (Erosion, Dilation, Opening, Closing) — MCQs | Digital Image Processing

  30. Boundary Extraction — MCQs | Digital Image Processing

  31. Skeletonization — MCQs | Digital Image Processing

  32. Connected Components Labeling — MCQs | Digital Image Processing

  33. Texture Analysis (GLCM, LBP, Gabor Filters) — MCQs | Digital Image Processing

  34. Shape Descriptors (Perimeter, Area, Compactness, Eccentricity) — MCQs | Digital Image Processing

  35. Statistical Features (Mean, Variance, Skewness) — MCQs | Digital Image Processing

  36. Principal Component Analysis (PCA) — MCQs | Digital Image Processing

  37. Linear Discriminant Analysis (LDA) — MCQs | Digital Image Processing

  38. Feature Matching (SIFT, SURF, ORB) — MCQs | Digital Image Processing

  39. Image Registration — MCQs | Digital Image Processing

  40. Image Stitching — MCQs | Digital Image Processing

  41. Motion Detection & Optical Flow — MCQs | Digital Image Processing

  42. Background Subtraction — MCQs | Digital Image Processing

  43. Object Detection & Tracking — MCQs | Digital Image Processing

  44. Template Matching — MCQs | Digital Image Processing

  45. Pattern Recognition (KNN, SVM, ANN) — MCQs | Digital Image Processing

  46. Image Classification — MCQs | Digital Image Processing

  47. Image Clustering — MCQs | Digital Image Processing

  48. Image Compression (RLE, Huffman, LZW, JPEG, JPEG2000) — MCQs | Digital Image Processing

  49. Video Compression (MPEG, H.264) — MCQs | Digital Image Processing

  50. Image Fusion (Pixel, Feature, Decision Level) — MCQs | Digital Image Processing

  51. Image Watermarking — MCQs | Digital Image Processing

  52. Steganography — MCQs | Digital Image Processing

  53. Face Detection & Recognition — MCQs | Digital Image Processing

  54. Gesture Recognition — MCQs | Digital Image Processing

  55. 3D Image Processing — MCQs | Digital Image Processing

  56. Stereo Vision & Depth Estimation — MCQs | Digital Image Processing

  57. Medical Image Analysis (CT, MRI, Ultrasound) — MCQs | Digital Image Processing

  58. Remote Sensing Image Processing — MCQs | Digital Image Processing

  59. Satellite Image Enhancement — MCQs | Digital Image Processing

  60. Deep Learning for Image Processing (CNN, GANs, Autoencoders) — MCQs | Digital Image Processing

  61. Image Captioning — MCQs | Digital Image Processing

  62. Semantic & Instance Segmentation (Mask R-CNN, U-Net) — MCQs | Digital Image Processing

  63. Super Resolution (SRCNN, ESRGAN) — MCQs | Digital Image Processing

  64. Image Inpainting — MCQs | Digital Image Processing

  65. Image Style Transfer — MCQs | Digital Image Processing

  66. Real-Time Image Processing — MCQs | Digital Image Processing

  67. Augmented Reality (AR) & Virtual Reality (VR) — MCQs | Digital Image Processing

  68. DIP using MATLAB/OpenCV/Python — MCQs | Digital Image Processing

  69. DIP in IoT & Embedded Systems — MCQs | Digital Image Processing

  70. Ethics & Privacy in Image Processing — MCQs | Digital Image Processing

Computer Science Repeated MCQs Book Download

Exit mobile version