[Paper Review] A Novel Improved Mask RCNN for Multiple Targets Detection in the Indoor Complex Scenes

Zongmin Liu, Jirui Wang+3 more|arXiv (Cornell University)|Jan 7, 2023

Advanced Neural Network Applications4 citations

TL;DR

This paper proposes an improved Mask R-CNN model enhanced with a Convolutional Block Attention Module (CBAM) to improve multiple object detection in complex indoor environments. By integrating channel and spatial attention mechanisms, the model achieves higher accuracy and stronger anti-interference performance compared to baseline methods, with detection speed maintained at near-real-time levels.

ABSTRACT

With the expansive aging of global population, service robot with living assistance applied in indoor scenes will serve as a crucial role in the field of elderly care and health in the future. Service robots need to detect multiple targets when completing auxiliary tasks. However, indoor scenes are usually complex and there are many types of interference factors, leading to great challenges in the multiple targets detection. To overcome this technical difficulty, a novel improved Mask RCNN method for multiple targets detection in the indoor complex scenes is proposed in this paper. The improved model utilizes Mask RCNN as the network framework. On this basis, Convolutional Block Attention Module (CBAM) with channel mechanism and space mechanism is integrated, and the influence of different background, distance, angle and interference factors are comprehensively considered. Meanwhile, in order to evaluate the detection and identification effects of the established model, a comprehensive evaluation system based on loss function and Mean Average Precision (mAP) is established. For verification, experiments on the detection and identification effects under different distances, backgrounds, angles and interference factors were conducted. The results show that designed model improves the accuracy to a higher level and has a better anti-interference ability than other methods when the detection speed was nearly the same.

Motivation & Objective

• To address the challenge of accurate multiple target detection in complex indoor environments with high interference.
• To improve detection robustness under varying distances, angles, and background conditions commonly found in indoor settings.
• To enhance feature representation in deep networks by integrating attention mechanisms into the Mask R-CNN framework.
• To develop a comprehensive evaluation system using mAP and loss function for model validation.
• To validate the model’s superiority in real-world indoor scenarios through controlled experiments.

Proposed method

• The proposed method extends the standard Mask R-CNN architecture by integrating a Convolutional Block Attention Module (CBAM) that jointly models channel and spatial attention.
• CBAM enhances feature learning by recalibrating feature maps based on channel-wise and spatial-wise importance, improving representation under complex conditions.
• The model is trained using a multi-object detection loss function combining classification, bounding box regression, and mask loss for end-to-end optimization.
• A comprehensive evaluation system is established using Mean Average Precision (mAP) and loss function to assess performance across diverse scenarios.
• Experiments are conducted under controlled variations in distance, background complexity, viewpoint angles, and environmental interference to test robustness.
• The model is trained and evaluated on a custom or standard indoor dataset, though the exact dataset is not specified in the abstract.

Experimental results

Research questions

• RQ1 How does the integration of CBAM into Mask R-CNN affect detection accuracy in complex indoor scenes?
• RQ2 To what extent does the improved model maintain performance under varying distances and angles?
• RQ3 How does the model compare to baseline methods in terms of robustness to background clutter and visual interference?
• RQ4 What is the trade-off between detection accuracy and inference speed in the proposed architecture?
• RQ5 Can the proposed evaluation system effectively capture performance differences across diverse indoor conditions?

Key findings

• The proposed model achieves higher detection accuracy than standard Mask R-CNN and other baseline methods in complex indoor scenes.
• The integration of CBAM significantly improves the model’s ability to handle interference from background clutter and occlusions.
• The model maintains near-real-time inference speed while achieving superior mAP scores across test conditions.
• The model demonstrates consistent performance improvements across varying distances and viewing angles, indicating strong generalization.
• The comprehensive evaluation system effectively quantifies performance gains, especially in challenging scenarios with high visual noise.
• The results confirm that attention mechanisms enhance feature representation, leading to better detection and segmentation of multiple targets.