Convolutional neural network (CNN) architectures are generally heavy on memory and computational requirements which make them infeasible for embedded systems with limited hardware resources. We propose dual convolutional kernels (DualConv) for constructing lightweight deep neural networks. DualConv combines $3\times 3$ and $1\times 1$ convolutional kernels to process the same input feature map channels simultaneously and exploits the group convolution technique to efficiently arrange convolutional filters. DualConv can be employed in any CNN model such as VGG-16 and ResNet-50 for image classification, you only look once (YOLO) and R-CNN for object detection, or fully convolutional network (FCN) for semantic segmentation. In this work, we extensively test DualConv for classification since these network architectures form the backbone for many other tasks. We also test DualConv for image detection on YOLO-V3. Experimental results show that, combined with our structural innovations, DualConv significantly reduces the computational cost and number of parameters of deep neural networks while surprisingly achieving slightly higher accuracy than the original models in some cases. We use DualConv to further reduce the number of parameters of the lightweight MobileNetV2 by 54% with only 0.68% drop in accuracy on CIFAR-100 dataset. When the number of parameters is not an issue, DualConv increases the accuracy of MobileNetV1 by 4.11% on the same dataset. Furthermore, DualConv significantly improves the YOLO-V3 object detection speed and improves its accuracy by 4.4% on PASCAL visual object classes (VOC) dataset.
1 March 2020 in Computer Methods and Programs in Biomedicine
DEtection TRansformer (DETR) for object detection reaches competitive performance compared with Faster R-CNN via a transformer encoder-decoder architecture. However, trained with scratch transformers, DETR needs large-scale training data and an extreme long training schedule even on COCO dataset. Inspired by the great success of pre-training transformers in natural language processing, we propose a novel pretext task named random query patch detection in Unsupervised Pre-training DETR (UP-DETR). Specifically, we randomly crop patches from the given image and then feed them as queries to the decoder. The model is pre-trained to detect these query patches from the input image. During the pre-training, we address two critical issues: multi-task learning and multi-query localization. (1) To trade off classification and localization preferences in the pretext task, we find that freezing the CNN backbone is the prerequisite for the success of pre-training transformers. (2) To perform multi-query localization, we develop UP-DETR with multi-query patch detection with attention mask. Besides, UP-DETR also provides a unified perspective for fine-tuning object detection and one-shot detection tasks. In our experiments, UP-DETR significantly boosts the performance of DETR with faster convergence and higher average precision on object detection, one-shot detection and panoptic segmentation. Code and pre-training models: https://github.com/dddzg/up-detr .
31 July 2023 in IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
Ultrasound computed tomography (USCT) is a fast-emerging imaging modality that is expected to help detect and characterize breast tumors by quantifying the distribution of the speed of sound (SOS) and acoustic attenuation in breast tissue. High-quality quantitative SOS reconstruction in USCT requires a large number of transducers, which incurs high system costs and slow computation. In contrast, sparsely distributed arrays are low-cost and fast but significantly degrade image quality. Thus, we propose a framework to achieve high-quality SOS Reconstruction under Sparse Sampling based on a convolutional neural Network (SRSS-Net) with faster computation. We first apply the bent-ray algorithm to sparsely sampled data and then apply the SRSS-Net to efficiently improve the image quality. Experimental results on synthetic and real datasets demonstrate that the proposed SRSS-Net provides reconstructions that are superior to those of state-of-the-art methods in terms of artifact suppression, structural preservation, quantitative restoration, and computational speed. As demonstrated in our experiments, the fine-tuning training strategy is suggested when applying SRSS-Net to real-world circumstances. The imaging and computational performance of SRSS-Net on the inhomogeneous breast phantom further demonstrates that SRSS-Net has great potential in real-time breast cancer detection.
10 November 2023 in Biomedical Signal Processing and Control