Multi-scale Detection Network Research Articles

Detecting the Background-Similar Objects in Complex Transportation Scenes

With the development of intelligent transportation systems, most human objects can be accurately detected in normal road scenes. However, the detection accuracy usually decreases sharply when the pedestrians are merged into the background with very similar colors or textures. In this paper, a camouflaged object detection method is proposed to detect the pedestrians or vehicles from the highly similar background. Specifically, we design a guide-learning-based multi-scale detection network (GLNet) to distinguish the weak semantic distinction between the pedestrian and its similar background, and output an accurate segmentation map to the autonomous driving system. The proposed GLNet mainly consists of a backbone network for basic feature extraction, a guide-learning module (GLM) to generate the principal prediction map, and a multi-scale feature enhancement module (MFEM) for prediction map refinement. Based on the guide learning and coarse-to-fine strategy, the final prediction map can be obtained with the proposed GLNet which precisely describes the position and contour information of the pedestrians or vehicles. Extensive experiments on four benchmark datasets, e.g., CHAMELEON, CAMO, COD10K, and NC4K, demonstrate the superiority of the proposed GLNet compared with several existing state-of-the-art methods.

RACNet: risk assessment Net of cervical lesions in colposcopic images

ABSTRACT In colposcopy-assisted diagnosis, the difference between the different lesion grades of colposcopic images is small, and the visual similarity is high. Therefore, it is a very challenging task to accurately diagnose cervical lesions through colposcopic images. This paper proposes a new risk assessment net of cervical lesions in colposcopic images (RACNet). The RACNet mainly consists of two parts. At first, the location and grade of lesions in different scales are detected through a multi-scale detection network. Then, these lesions are classified by designing a multi-branch convolutional neural network (CNN) to improve the performance of risk assessment. The RACNet was compared with the most advanced methods and colposcopists under the same condition. The experimental results show that the RACNet in this paper is superior to other methods, with an accuracy rate of 84.5%, which is 15% higher than the average level of colposcopists. It can provide clinicians with auxiliary diagnosis and reduce missed diagnosis and misdiagnosis.

A deep learning-based method for detecting and classifying the ultrasound images of suspicious thyroid nodules.

The incidence of thyroid cancer has significantly increased in the last few decades. However, diagnosis of the thyroid nodules is labor and time intensive for radiologists and strongly depends on the personal experience of the radiologists. In this pursuit, the present study envisaged to develop a deep learning-based computer-aided diagnosis (CAD) method that enabled the automatic detection and classification of suspicious thyroid nodules in order to reduce the unnecessary fine-needle aspiration biopsy. The CAD method consisted of two main parts: detecting the location of thyroid nodules using a multiscale detection network and classifying the detected thyroid nodules by an attention-based classification network. The performance of the proposed method was evaluated and compared with that of other state-of-the-art deep learning methods and experienced radiologists. The proposed detection method outperformed three other detection architectures (average precision, 82.1%vs. 78.3%, 77.2%, and 74.8%). Moreover, the classification method showed a superior performance compared with four other state-of-the-art classification networks (accuracy, 94.8%vs. 91.2%, 85.0%, 80.8%, and 72.1%) and that by experienced radiologists (mean value of area under the curve, 0.941vs. 0.833). Our study verified the high efficiency of the proposed detection method. The findings can help improve the diagnostic performance of radiologists. However, the developed CAD system requires more training and evaluation in a large-population study.

Aircraft Detection for Remote Sensing Images Based on Deep Convolutional Neural Networks

Aircraft detection for remote sensing images, as one of the fields of computer vision, is one of the significant tasks of image processing based on deep learning. Recently, many high-performance algorithms for aircraft detection have been developed and applied in different scenarios. However, the proposed algorithms still have a series of problems; for instance, the algorithms will miss some small-scale aircrafts when applied to the remote sensing image. There are two main reasons for the problem; one reason is that the aircrafts in the remote sensing image are usually small in size, leading to detecting difficulty. The other reason is that the background of the remote sensing image is usually complex, so the algorithms applied to the scenario are easy to be affected by the background. To address the problem of small size, this paper proposes the Multiscale Detection Network (MSDN) which introduces a multiscale detection architecture to detect small-scale aircrafts. With the intention to resist the background noise, this paper proposes the Deeper and Wider Module (DAWM) which increases the perceptual field of the network to alleviate the affection. Besides, to address the two problems simultaneously, this paper introduces the DAWM into the MSDN and names the novel network structure as Multiscale Refined Detection Network (MSRDN). The experimental results show that the MSRDN method has detected the small-scale aircrafts that other algorithms missed and the performance indicators have higher performance than other algorithms.

Real-time object detection algorithm based on improved YOLOv3

Object detection is a challenging computer vision problem with numerous practical applications. Due to low accuracy and slow detection speed in object detection, we propose a real-time object detection algorithm based on YOLOv3. First, to solve the problem that features are likely to be lost in the feature extraction process of YOLOv3, a DB-Darknet-53 feature extraction network embedded in inception structure is designed, which effectively reduces the loss of features. Second, the detection network of YOLOv3 and the reuse of deep features in multiscale detection network are improved. Finally, the numbers and sizes of anchor boxes are selected by K -means clustering analysis, and the detection model is obtained by means of multiscale training. The improved algorithm has a mean average precision of 0.835 on the PASCAL VOC data set and a detection speed of 35.8 f / s, which is better than YOLOv3.