Monocular 3D object detection has recently become prevalent in autonomous driving and navigation applications due to its cost-efficiency and easy-to-embed to existent vehicles. The most challenging task in monocular vision is to estimate a reliable object’s location cause of the lack of depth information in RGB images. Many methods tackle this ill-posed problem by directly regressing the object’s depth or take the depth map as a supplement input to enhance the model’s results. However, the performance relies heavily on the estimated depth map quality, which is bias to the training data. In this work, we propose depth-adaptive convolution to replace the traditional 2D convolution to deal with the divergent context of the image’s features. This lead to significant improvement in both training convergence and testing accuracy. Second, we propose a ground plane model that utilizes geometric constraints in the pose estimation process. With the new method, named GAC3D, we achieve better detection results. We demonstrate our approach on the KITTI 3D Object Detection benchmark, which outperforms existing monocular methods.

Attention is all we need as long as we have enough data. Even so, it is sometimes not easy to determine how much data is enough while the models are becoming larger and larger. In this paper, we propose HYDRA heads, lightweight pretrained linguistic self-attention heads to inject knowledge into transformer models without pretraining them again. Our approach is a balanced paradigm between leaving the models to learn unsupervised and forcing them to conform to linguistic knowledge rigidly as suggested in previous studies. Our experiment proves that the approach is not only the boost performance of the model but also lightweight and architecture friendly. We empirically verify our framework on benchmark datasets to show the contribution of linguistic knowledge to a transformer model. This is a promising result for a new approach to transferring knowledge from linguistic resources into transformer-based models.

No abstract

Many alternative approaches for 3D object detection using a singular camera have been studied instead of leveraging high-precision 3D LiDAR sensors incurring a prohibitive cost. Recently, we proposed a novel approach for 3D object detection by employing a ground plane model that utilizes geometric constraints named GAC3D to improve the results of the deep-based detector. GAC3D adopts an adaptive depth convolution to replace the traditional 2D convolution to deal with the divergent context of the image’s feature, leading to a significant improvement in both training convergence and testing accuracy on the KITTI 3D object detection benchmark. This article presents an alternative architecture named eGAC3D that adopts a revised depth adaptive convolution with variant guidance to improve detection accuracy. Additionally, eGAC3D utilizes the pixel adaptive convolution to leverage the depth map to guide our model for detection heads instead of using an external depth estimator like other methods leading to a significant reduction of time inference. The experimental results on the KITTI benchmark show that our eGAC3D outperforms not only our previous GAC3D but also many existing monocular methods in terms of accuracy and inference time. Moreover, we deployed and optimized the proposed eGAC3D framework on an embedded platform with a low-cost GPU. To the best of the authors’ knowledge, we are the first to develop a monocular 3D detection framework on embedded devices. The experimental results on Jetson Xavier NX demonstrate that our proposed method can achieve nearly real-time performance with appropriate accuracy even with the modest hardware resource.