Performance Analysis of Traditional VQA Models Under Limited Computational Resources
Jihao Gu
TL;DR
This work analyzes Visual Question Answering under resource constraints by comparing traditional sequential and CNN-based models, with a focus on numerical and counting questions. It finds that a BidGRU-based Question Encoding with embedding $d=300$ and vocabulary size $V=3000$ delivers the best balance between accuracy and efficiency, especially when paired with effective attention and a counting module. Key contributions include a thorough ablation of components (attention, counting, fusion) and a demonstration that carefully tuned traditional architectures can approach or match more complex models on constrained hardware. The findings offer practical guidance for deploying efficient VQA systems in environments with limited computational capacity, such as medical or industrial settings.
Abstract
In real-world applications where computational resources are limited, effectively integrating visual and textual information for Visual Question Answering (VQA) presents significant challenges. This paper investigates the performance of traditional models under computational constraints, focusing on enhancing VQA performance, particularly for numerical and counting questions. We evaluate models based on Bidirectional GRU (BidGRU), GRU, Bidirectional LSTM (BidLSTM), and Convolutional Neural Networks (CNN), analyzing the impact of different vocabulary sizes, fine-tuning strategies, and embedding dimensions. Experimental results show that the BidGRU model with an embedding dimension of 300 and a vocabulary size of 3000 achieves the best overall performance without the computational overhead of larger models. Ablation studies emphasize the importance of attention mechanisms and counting information in handling complex reasoning tasks under resource limitations. Our research provides valuable insights for developing more efficient VQA models suitable for deployment in environments with limited computational capacity.