Tidying Deep Saliency Prediction Architectures
This work addresses the need for more efficient and interpretable saliency models for applications in robotics and computer vision, though it is incremental in simplifying existing architectures.
The paper tackles the problem of overly complex deep saliency prediction models by identifying four key components and proposing simpler alternatives, resulting in two novel architectures (SimpleNet and MDNSal) that achieve state-of-the-art performance on public benchmarks, such as notable gains on SALICON, and can run at 25fps for real-time applications.
Learning computational models for visual attention (saliency estimation) is an effort to inch machines/robots closer to human visual cognitive abilities. Data-driven efforts have dominated the landscape since the introduction of deep neural network architectures. In deep learning research, the choices in architecture design are often empirical and frequently lead to more complex models than necessary. The complexity, in turn, hinders the application requirements. In this paper, we identify four key components of saliency models, i.e., input features, multi-level integration, readout architecture, and loss functions. We review the existing state of the art models on these four components and propose novel and simpler alternatives. As a result, we propose two novel end-to-end architectures called SimpleNet and MDNSal, which are neater, minimal, more interpretable and achieve state of the art performance on public saliency benchmarks. SimpleNet is an optimized encoder-decoder architecture and brings notable performance gains on the SALICON dataset (the largest saliency benchmark). MDNSal is a parametric model that directly predicts parameters of a GMM distribution and is aimed to bring more interpretability to the prediction maps. The proposed saliency models can be inferred at 25fps, making them suitable for real-time applications. Code and pre-trained models are available at https://github.com/samyak0210/saliency.