Abstract:Binary Neural Network (BNN) exhibits substantial potential for low-cost and energy-efficient deployment, notably in the application of BNN-facilitated object detection on embedded devices. Despite its high compression ratio, the optimization of BNN encounters a significant challenge due to the discrete nature of binary weights and activations. Achieving good generalization performance poses a notable obstacle in this regard. Previous BNN training schemes that only focus on minimizing the empirical loss can easily suffer from overfitting problem, yielding poor generalization capacity when being exposed to unseen data. Sharpness-Aware Minimization (SAM), which simultaneously minimizes the loss value and sharpness of the loss landscape, has emerged as an effective approach for enhancing the generalization ability of neural networks, and has been demonstrated to be effective in improving the performance of BNNs designed for classification tasks. However, their work does not address the optimization challenge of applying SAM to models that handle multiple tasks, such as binary object detection, which involves both classification and location tasks. To address this issue, we propose a modified SAM scheme, denoted as BNN-SAM, which introduces a new objective allowing for the direct calculation of the optimal update vector. Moreover, the proposed scheme fosters multi-task optimization by establishing a common global flat minima for all the concerned tasks. This attribute renders it particularly fitting for scenarios such as object detection, inherently necessitating the joint optimization of both classification and localization. Comprehensive experiments on the PASCAL VOC and MS COCO datasets have shown that BNN-SAM can easily improve the performance of a baseline binary SSD300 detector to outperform state-of-the-art binary detectors, including BiDet, AutoBiDet and LWS-Det, by 6.5%, 5.0%, 1.1%, respectively, without the need of extra optimization approaches. Code is available at https://github.com/Anonymous2740/BNN-SAM.

Pixel-wise binary classification network for salient object detection

SAFPN: a Full Semantic Feature Pyramid Network for Object Detection

Improving object detection with deep convolutional networks via Bayesian optimization and structured prediction

Salient Object Detection Based on Visual Perceptual Saturation and Two-Stream Hybrid Networks.

Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection.

Accurate salient object detection via dense recurrent connections and residual-based hierarchical feature integration.

Salient Object Detection Via Multi-Scale Neural Network.

Contrast-Oriented Deep Neural Networks for Salient Object Detection

Deep Contrast Learning for Salient Object Detection

BASNet: Boundary-Aware Salient Object Detection

HFENet: Hybrid feature encoder network for detecting salient objects in RGB-thermal images

DHSNet: Deep Hierarchical Saliency Network for Salient Object Detection

AWANet: Attentive-Aware Wide-Kernels Asymmetrical Network with Blended Contour Information for Salient Object Detection

PDNet: Prior-model Guided Depth-enhanced Network for Salient Object Detection

Hierarchical Dynamic Filtering Network for RGB-D Salient Object Detection

Fcn And Unit-Linking Pcnn Based Image Saliency Detection

BNN-SAM: Improving generalization of binary object detector by Seeking Flat Minima

Bi-DAINet: Bi-Directional Discard-Accept-Integrate Network for salient object detection

Salient Object Detection with Pyramid Attention and Salient Edges

A Pooling-Based Feature Pyramid Network for Salient Object Detection

Attention and boundary guided salient object detection