Question Bank
Computer Vision interview questions
An image model may perform well on a test set and fail when the camera or lighting changes. Review architectures, data preparation and evaluation. For each choice, discuss which changes in the input could make the solution unreliable.
Showing 20 of 75 questions
What is a convolution and how is it used in convolutional neural networks?
Explain what overfitting is and how it can be avoided in the context of training deep learning models.
What is image augmentation and what types of augmentation do you know?
How does the attention mechanism work in transformers and what impact does it have on model performance?
Explain what NMS (Non-Maximum Suppression) is and why it is important in object detection tasks.
What is a GAN and what are the main components of a GAN model?
Why is it important to use pre-trained models and what are the advantages of fine-tuning compared to training from scratch?
What is the difference between FCN and U-Net in semantic segmentation tasks?
What are the main metrics for evaluating the quality of object detection models and how are they calculated?
What are convolutions and how are they used in computer vision?
Explain what data augmentation is and why it is needed.
What is IoU and how is it used in object detection tasks?
Compare the VGG and ResNet architectures. What are their main differences?
When and why should transfer learning be used?
Explain how the Mask R-CNN method works for semantic segmentation.
What are the approaches to prevent mode collapse in GANs?
Explain how deep diffusion models (DDPM) work.
How would you explain the differences between SSD and Faster R-CNN in object detection?
What are activation functions in convolutional neural networks and why are they important?
Explain what data augmentation is and what role it plays in training computer vision models.
Prepare for your next interview with Vibe Interview.
Download Vibe Interview