Consider a scenario where you are developing a virtual assistant that can answer questions about images. You have a large dataset of images and corresponding question-answer pairs. Which architecture is BEST suited for this task?

• A.A convolutional neural network (CNN) for image feature extraction followed by a recurrent neural network (RNN) for question encoding and a fully connected layer for answer prediction.
• B.A transformer-based model that processes both images and questions as sequences of tokens, allowing for attention-based interaction between modalities.
• C.A simple feedforward neural network that takes flattened image pixels and question embeddings as input.
• D.A combination of a pre-trained word embedding model (e.g., Word2Vec) for the question and a separate model for image classification.
• E.A Support Vector Machine (SVM) trained on image features and question keywords.

Comment: *

Name: *

Email: *

Verification: *

You have a multimodal model that processes images and text, and you want to deploy it on an edge device with limited computational resources. Which of the following hardware acceleration strategies would be MOST effective in improving the model's inference speed on the edge device?

• A.Using NVIDIA TensorRT to optimize the model for the specific edge device.
• B.Implementing distributed inference across multiple edge devices.
• C.Using a larger batch size to improve GPU utilization.
• D.Converting the model to a smaller architecture with fewer parameters, accepting a lower accuracy.
• E.Offloading complex computations to a cloud server.

Comment: *

Name: *

Email: *

Verification: *

You are building a system to generate captions for images. You want to evaluate how well the generated captions describe the content of the images. Which of the following metrics are most suitable for evaluating the quality of image captions?

• A.Pixel Accuracy.
• B.BLEU (Bilingual Evaluation Understudy).
• C.Inception Score.
• D.ROUGE (Recall-Oriented Understudy for Gisting Evaluation).
• E.F 1-Score

Comment: *

Name: *

Email: *

Verification: *

You have a dataset containing information about sales performance for different regions in the last ten years.
Which type of data visualization would be most appropriate to compare the sales performance across regions on a year-by-year basis?

• A.Scatter plot
• B.Line chart
• C.Bar chart
• D.Pie chart

Comment: *

Name: *

Email: *

Verification: *

You are building a text-to-image generation pipeline using CLIP and a diffusion model. After training, you notice that the generated images often lack the specific details mentioned in the text prompts. Which of the following strategies could you employ to improve the alignment between text and image?

• A.Increase the number of layers in the I-I-Net architecture of the diffusion model.
• B.Fine-tune the CLIP model on a dataset of text-image pairs relevant to your desired domain.
• C.Use negative prompt engineering to guide the diffusion process away from undesired attributes.
• D.Increase the number of diffusion steps during the image generation process.
• E.All of the above.

Comment: *

Name: *

Email: *

Verification: *