LPNet: Retina Inspired Neural Network for Object Detection and Recognition

The detection of rotated objects is a meaningful and challenging research work. Although the state-of-the-art deep learning models have feature invariance, especially convolutional neural networks (CNNs), their architectures did not specifically design for rotation invariance. They only slightly compensate for this feature through pooling layers. In this study, we propose a novel network, named LPNet, to solve the problem of object rotation. LPNet improves the detection accuracy by combining retina-like log-polar transformation. Furthermore, LPNet is a plug-and-play architecture for object detection and recognition. It consists of two parts, which we name as encoder and decoder. An encoder extracts images which feature in log-polar coordinates while a decoder eliminates image noise in cartesian coordinates. Moreover, according to the movement of center points, LPNet has stable and sliding modes. LPNet takes the single-shot multibox detector (SSD) network as the baseline network and the visual geometry group (VGG16) as the feature extraction backbone network. The experiment results show that, compared with conventional SSD networks, the mean average precision (mAP) of LPNet increased by 3.4% for regular objects and by 17.6% for rotated objects.

Predicting Arbitrary-Oriented Objects as Points in Remote Sensing Images

To detect rotated objects in remote sensing images, researchers have proposed a series of arbitrary-oriented object detection methods, which place multiple anchors with different angles, scales, and aspect ratios on the images. However, a major difference between remote sensing images and natural images is the small probability of overlap between objects in the same category, so the anchor-based design can introduce much redundancy during the detection process. In this paper, we convert the detection problem to a center point prediction problem, where the pre-defined anchors can be discarded. By directly predicting the center point, orientation, and corresponding height and width of the object, our methods can simplify the design of the model and reduce the computations related to anchors. In order to further fuse the multi-level features and get accurate object centers, a deformable feature pyramid network is proposed, to detect objects under complex backgrounds and various orientations of rotated objects. Experiments and analysis on two remote sensing datasets, DOTA and HRSC2016, demonstrate the effectiveness of our approach. Our best model, equipped with Deformable-FPN, achieved 74.75% mAP on DOTA and 96.59% on HRSC2016 with a single-stage model, single-scale training, and testing. By detecting arbitrarily oriented objects from their centers, the proposed model performs competitively against oriented anchor-based methods.

Rotation Invariant Networks for Image Classification for HPC and Embedded Systems

Convolutional Neural Network (CNNs) models’ size reduction has recently gained interest due to several advantages: energy cost reduction, embedded devices, and multi-core interfaces. One possible way to achieve model reduction is the usage of Rotation-invariant Convolutional Neural Networks because of the possibility of avoiding data augmentation techniques. In this work, we present the next step to obtain a general solution to endowing CNN architectures with the capability of classifying rotated objects and predicting the rotation angle without data-augmentation techniques. The principle consists of the concatenation of a representation mapping transforming rotation to translation and a shared weights predictor. This solution has the advantage of admitting different combinations of various basic, existing blocks. We present results obtained using a Gabor-filter bank and a ResNet feature backbone compared to previous other solutions. We also present the possibility to select between parallelizing the network in several threads for energy-aware High Performance Computing (HPC) applications or reducing the memory footprint for embedded systems. We obtain a competitive error rate on classifying rotated MNIST and outperform existing state-of-the-art results on CIFAR-10 when trained on up-right examples and validated on random orientations.

The Fast Integration of a Rotated Rectangle Applied to the Rotated Haar-Like Features for Rotated Objects Detection

In the area of Object Detection, the most important step is the extraction of object features. One of the most used approaches is HaarLike features and the Integral Image technique to integrate them. The Integral Image technique, used by Viola and Jones, is generally used to calculate the integral of a rectangular filter in an input picture. This filter is a rectilinear rectangle. We propose a method to integrate a rotated one by any angle of rotation inside an image based on the Bresenham algorithm of drawing a segment. We use some pixels – called key points - that forms the four segments of a rotated rectangle, to calculate its Integral Image. Our method focuses on three essential tasks; the first is to determine the rule for drawing a segment (SDR), the second is to identify all the key points of the rectangle r, and the third is to calculate the integral image. The speed of this method depends on the size and angle of rotation of the rectangle. To demonstrate the efficiency of our idea, we applied it to the rotated Haar-like features that we proposed in a later work [12], which had as objectives the improvement of the Viola and Jones algorithm to detect the rotated faces in a given image. We performed tests on more widespread databases of images, which showed that the application of this technique to rotated Haar-Like features improves the performance of object detectors, in general, and faces in particular.