TO–YOLOX: a pure CNN tiny object detection model for remotesensing images

Remote sensing and deep learning are being widely combined in tasks such as urban planning and disaster prevention.However,due to interference occasioned by density,overlap,and coverage,the tiny object detection in remote sensing images has always been a difficult problem.Therefore,we propose a novel TO–YOLOX(Tiny Object–You Only Look Once)model.TO–YOLOX possesses a MiSo(Multiple-in-Singleout)feature fusion structure,which exhibits a spatial-shift structure,and the model balances positive and negative samples and enhances the information interaction pertaining to the local patch of remote sensing images.TO–YOLOX utilizes an adaptive IOU-T(Intersection Over Uni-Tiny)loss to enhance the localization accuracy of tiny objects,and it applies attention mechanism Group-CBAM(group-convolutional block attention module)to enhance the perception of tiny objects in remote sensing images.To verify the effectiveness and efficiency of TO–YOLOX,we utilized three aerial-photography tiny object detection datasets,namely VisDrone2021,Tiny Person,and DOTA–HBB,and the following mean average precision(mAP)values were recorded,respectively:45.31%(+10.03%),28.9%(+9.36%),and 63.02%(+9.62%).With respect to recognizing tiny objects,TO–YOLOX exhibits a stronger ability compared with Faster R-CNN,RetinaNet,YOLOv5,YOLOv6,YOLOv7,and YOLOX,and the proposed model exhibits fast computation.

Real-time detection network for tiny traffic sign using multi-scale attention module

As one of the key technologies of intelligent vehicles, traffic sign detection is still a challenging task because of the tiny size of its target object. To address the challenge, we present a novel detection network improved from yolo-v3 for the tiny traffic sign with high precision in real-time. First, a visual multi-scale attention module(MSAM), a light-weight yet effective module, is devised to fuse the multi-scale feature maps with channel weights and spatial masks. It increases the representation power of the network by emphasizing useful features and suppressing unnecessary ones. Second, we exploit effectively fine-grained features about tiny objects from the shallower layers through modifying backbone Darknet-53 and adding one prediction head to yolo-v3. Finally, a receptive field block is added into the neck of the network to broaden the receptive field. Experiments prove the effectiveness of our network in both quantitative and qualitative aspects. The m AP@0.5 of our network reaches 0.965 and its detection speed is55.56 FPS for 512 × 512 images on the challenging Tsinghua-Tencent 100 k(TT100 k) dataset.