Towards Rotation Invariance in Object Detection

Kalra, Agastya; Stoppi, Guy; Brown, Bradley; Agarwal, Rishi; Kadambi, Achuta

doi:10.1109/iccv48922.2021.00351

Cited by 12 publications

(2 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The basic idea behind this improvement is that reducing the rotation invariance constraints on the model benefits detection. CNNs are not rotation invariant by default, and existing rotation augmentations for object detection do not always improve the rotation invariance of a model (Kalra et al, 2021). Additionally, as shown in Figure 11, stabilization enables the detector to recognize objects more effectively.…”

Section: Resultsmentioning

confidence: 99%

Multisensor fusion‐based maritime ship object detection method for autonomous surface vehicles

Zhang,

Shan,

Zhang

et al. 2023

Journal of Field Robotics

View full text Add to dashboard Cite

Autonomous surface vehicles face the challenge of accurately detecting nearby ships in the complex and ever‐changing maritime environment, which is vastly different from land areas. To address this issue, we propose an image‐based multisensor fusion object detection method that combines Light Detection and Rangings and cameras. Since point clouds have poor semantics, our method primarily relies on images, with point clouds used to support image detection. Our image detection scheme employs a tracking‐assisted detection method that leverages historical information to compensate for possible detection failures. Additionally, we designed a confidence‐association‐based fusion strategy to determine the final targets among the candidates. We conducted field experiments in an open‐sea area to demonstrate the accuracy and robustness of our method. The results of these experiments showed that our method is highly accurate and robust in challenging maritime scenarios. Our code and data set will be released on https://github.com/flakeice/mssd.

show abstract

Section: Resultsmentioning

confidence: 99%

Multisensor fusion‐based maritime ship object detection method for autonomous surface vehicles

Zhang,

Shan,

Zhang

et al. 2023

Journal of Field Robotics

View full text Add to dashboard Cite

show abstract

“…Rotations are the best way of augmentation, which creates the reliable synthetic images by just rotating the image objects without disturbing them. According to Kalra et al [13], the augmentation with rotation not only increases the dataset size, but also improves the trained model efficiency in classification and segmentation.…”

Section: D-cu-net and Kidney Tumor Segmentationmentioning

confidence: 99%

Depth Invariant 3D-CU-Net Model with Completely Connected Dense Skip Networks for MRI Kidney Tumor Segmentation

Parvathi¹,

Chandana²,

Harikiran³

2023

View full text Add to dashboard Cite

Due to the impact and importance of the kidney objects in human body, the kidney tumor analysis from three dimensional CT and MRI medical images becomes a pivotal research topic, which helps in diagnosing the kidney diseases like kidney stones, polycystic and kidney tumors etc. In deep learning, U-Net became a prominent and reliable solution for kidney image analysis and objects segmentation process. Although several research works were focused on kidney object detection and tumor segmentation from medical images, they are suffering from some intrinsic limitations due to: variance in network depths, enforced feature fusion, segmentation errors and inaccuracy. In order to address these limitations in kidney tumor segmentation process, in this paper we proposed the 3D-CU-Net model for kidney tumor segmentation, which is a custom variant of the U-Net. In 3D-CU-Net, the encoder-decoder network model is unified to tolerate the depth invariance issues, while training various input images with the same model. Completely connected dense skip connections are designed at each layer of 3D-CU-Net, to control the enforced feature fusion and to extract the crucial features. An integrated loss function is designed with Binary Cross Entropy (BCE) and Soft-Dice Coefficient (SDC) to mitigate the segmentation errors and inaccuracy. Experiments on TCGA-KIRC dataset with 3D-CU-NET recorded the high accuracy in kidney tumor segmentation with mIoU (91.21%) and mDSC (92.69%).

show abstract