# DOTA-DOAI **Repository Path**: zhuang_shuo/DOTA-DOAI ## Basic Information - **Project Name**: DOTA-DOAI - **Description**: This repo is the codebase for our team to participate in DOTA related competitions, including rotation and horizontal detection. - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 1 - **Forks**: 0 - **Created**: 2021-03-31 - **Last Updated**: 2025-08-20 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # DOTA-DOAI ## Abstract This repo is the codebase for our team to participate in DOTA related competitions, including rotation and horizontal detection. We mainly use [FPN](https://arxiv.org/abs/1612.03144)-based two-stage detector, and it is completed by [YangXue](https://yangxue0827.github.io/) and [YangJirui](https://github.com/yangJirui). ## Performance ### DOTA1.0 (Task1) | Model | Backbone | Training data | Val data | mAP | Model Link | Tricks | lr schd | Data Augmentation | GPU | Image/GPU | Configs | |:------------:|:------------:|:---------:|:-----------:|:----------:|:-----------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:| | [FPN](https://arxiv.org/abs/1612.03144) (baseline) | ResNet50_v1 (600,800,1024)->800 | DOTA1.0 trainval | DOTA1.0 test | 69.35 | [model](https://drive.google.com/file/d/1QRxFAQ_Nqj3kqagc-NtMWKlxKjJnQFLo/view?usp=sharing) | No | 1x | No | 2X GeForce RTX 2080 Ti | 1 | cfgs_dota1.0_res50_v2.py | | FPN | ResNet50_v1d (600,800,1024)->800 | DOTA1.0 trainval | DOTA1.0 test | 70.87 | [model](https://drive.google.com/file/d/1mdvfNgIuagFQfddIV12yx9TWcRQ65YTf/view?usp=sharing) | [**+InLD**]() | 1x | No | 2X GeForce RTX 2080 Ti | 1 | cfgs_dota1.0_res50_v3.py | | FPN | **ResNet152_v1d (600,800,1024)->MS** | DOTA1.0 trainval | DOTA1.0 test | 76.20 (76.54) | [model](https://drive.google.com/file/d/16lZEttBN3asEDP7lIryvH-PByjxm6A7K/view?usp=sharing) | **ALL** | **2x** | **Yes** | 2X GeForce RTX 2080 Ti | 1 | cfgs_dota1.0_res152_v1.py | ### DOTA1.0 (Task2) | Model | Backbone | Training data | Val data | mAP | Model Link | Tricks | lr schd | Data Augmentation | GPU | Image/GPU | Configs | |:------------:|:------------:|:---------:|:-----------:|:----------:|:-----------:|:---------:|:---------:|:---------:|:---------:|:---------:|:---------:| | [FPN](https://arxiv.org/abs/1612.03144) (baseline) | ResNet50_v1 (600,800,1024)->800 | DOTA1.0 trainval | DOTA1.0 test | 76.03 | [model](https://drive.google.com/file/d/1aT_wvoV9ao2COZHOe6OfJglBbnFa8qqa/view?usp=sharing) | No | 1x | No | 2X Quadro RTX 8000 | 1 | cfgs_dota1.0_res50_v2.py | | FPN (memory consumption) | **ResNet152_v1d (600,800,1024)->MS** | DOTA1.0 trainval | DOTA1.0 test | 81.23 | [model](https://drive.google.com/file/d/1HaSU75llga_Em1O73Jp8ZwOoHPZtaVGj/view?usp=sharing) | **ALL** | **2x** | **Yes** | 2X Quadro RTX 8000 | 1 | cfgs_dota1.0_res152_v1.py | ### Visualization ![1](demo.jpg) ## Performance of published papers on DOTA datasets ### DOTA1.0 (Task1) | Model | Backbone | mAP | Paper Link | Code Link | Remark | Recommend | |:-----:|:--------:|:---:|:----------:|:---------:|:---------:|:---------:| | FR-O (DOTA) | ResNet101 | 52.93 | [CVPR2018](http://openaccess.thecvf.com/content_cvpr_2018/html/Xia_DOTA_A_Large-Scale_CVPR_2018_paper.html) | [MXNet](https://github.com/jessemelpolio/Faster_RCNN_for_DOTA) | DOTA dataset, baseline | :white_check_mark: | | IENet | ResNet101 | 57.14 | [arXiv:1912.00969](https://arxiv.org/abs/1912.00969) | - | anchor free | | | TOSO | ResNet101 | 57.52 | [ICASSP2020](https://ieeexplore.ieee.org/document/9053562) | - | geometric transformation | | | PIoU Loss | DLA-34 | 60.5 | [ECCV2020](https://arxiv.org/abs/2007.09584) | [Pytorch](https://github.com/clobotics/piou) | IoU loss, anchor free | :white_check_mark: | | R²CNN | ResNet101 | 60.67 | [arXiv:1706.09579](https://arxiv.org/abs/1706.09579) | [TF](https://github.com/DetectionTeamUCAS/R2CNN_Faster-RCNN_Tensorflow) | scene text, multi-task, different pooled sizes, baseline | :white_check_mark: | | RRPN | ResNet101 | 61.01 | [TMM](https://ieeexplore.ieee.org/document/8323240) [arXiv:1703.01086](https://arxiv.org/pdf/1703.01086.pdf) | [TF](https://github.com/DetectionTeamUCAS/RRPN_Faster-RCNN_Tensorflow) | scene text, rotation proposals, baseline | :white_check_mark: | | RetinaNet-H | ResNet101 | 64.73 | [arXiv:1908.05612](https://arxiv.org/abs/1908.05612) | [TF](https://github.com/DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation) | single stage, baseline | :white_check_mark: | | Axis Learning | ResNet101 | 65.98 | [Remote Sensing](https://www.mdpi.com/2072-4292/12/6/908) | - | single stage, anchor free | :white_check_mark: | | ICN | ResNet101 | 68.16 | [ACCV2018](https://link.springer.com/chapter/10.1007/978-3-030-20893-6_10) | - | image cascade, multi-scale | :white_check_mark: | | RADet | ResNeXt101 | 69.09 | [Remote Sensing](https://www.mdpi.com/2072-4292/12/3/389) | - | enhanced FPN, mask rcnn | | | RoI Transformer | ResNet101 | 69.56 | [CVPR2019](http://openaccess.thecvf.com/content_CVPR_2019/papers/Ding_Learning_RoI_Transformer_for_Oriented_Object_Detection_in_Aerial_Images_CVPR_2019_paper.pdf) | [MXNet](https://github.com/dingjiansw101/RoITransformer_DOTA), [Pytorch](https://github.com/dingjiansw101/AerialDetection) | roi transformer | :white_check_mark: | | P-RSDet | ResNet101 | 69.82 | [arXiv:2001.02988](https://arxiv.org/abs/2001.02988) | - | anchor free, polar coordinates | :white_check_mark: | | CAD-Net | ResNet101 | 69.90 | [TGRS](https://ieeexplore.ieee.org/document/8804364/) [arXiv:1903.00857](https://arxiv.org/abs/1903.00857) | - | attention | | | O²-DNet | Hourglass104 | 71.04 | [arXiv:1912.10694](https://arxiv.org/abs/1912.10694) | - | centernet, anchor free | :white_check_mark: | | AOOD | ResNet101 | 71.18 | [Neural Computing and Applications](https://link.springer.com/article/10.1007/s00521-020-04893-9) | - | attention + [R-DFPN](https://www.mdpi.com/2072-4292/10/1/132) | | | Cascade-FF | ResNet152 | 71.80 | [ICME2020](https://ieeexplore.ieee.org/abstract/document/9102807) | - | Refined RetinaNet + feature fusion | | | BBAVectors | ResNet101 | 72.32| [WACV2021](https://arxiv.org/abs/2008.07043) | [Pytorch](https://github.com/yijingru/BBAVectors-Oriented-Object-Detection) | keypoint based | :white_check_mark: | | SCRDet | ResNet101 | 72.61 | [ICCV2019](http://openaccess.thecvf.com/content_ICCV_2019/papers/Yang_SCRDet_Towards_More_Robust_Detection_for_Small_Cluttered_and_Rotated_ICCV_2019_paper.pdf) | [TF:R²CNN++](https://github.com/DetectionTeamUCAS/R2CNN-Plus-Plus_Tensorflow), IoU-Smooth L1: [RetinaNet-based](https://github.com/SJTU-Thinklab-Det/R3Det_Tensorflow), [R³Det-based](https://github.com/DetectionTeamUCAS/RetinaNet_Tensorflow_Rotation) | attention, angular boundary problem | :white_check_mark: | | SARD | ResNet101 | 72.95 | [Access](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8917630) | - | IoU-based weighted loss | | | GLS-Net | ResNet101 | 72.96 | [Remote Sensing](https://www.mdpi.com/2072-4292/12/9/1435) | - | attention, saliency pyramid | | | DRN | Hourglass104 | 73.23 | [CVPR(oral)](https://arxiv.org/abs/2005.09973) | [code](https://github.com/Anymake/DRN_CVPR2020) | centernet, feature selection module, dynamic refinement head, new dataset (SKU110K-R) | :white_check_mark: | | FADet | ResNet101 | 73.28 | [ICIP2019](https://ieeexplore.ieee.org/abstract/document/8803521) | - | attention | | | MFIAR-Net | ResNet152 | 73.49 | [Sensors](https://www.mdpi.com/1424-8220/20/6/1686/htm) | - | feature attention, enhanced FPN | | | R³Det | ResNet152 | 73.74 | [arXiv:1908.05612](https://arxiv.org/abs/1908.05612) | [TF](https://github.com/SJTU-Thinklab-Det/R3Det_Tensorflow), [Pytorch](https://github.com/SJTU-Thinklab-Det/r3det-on-mmdetection) | refined single stage, feature alignment | :white_check_mark: | | RSDet | ResNet152 | 74.10 | [arXiv:1911.08299](https://arxiv.org/abs/1911.08299) | - | quadrilateral bbox, angular boundary problem | :white_check_mark: | | Gliding Vertex | ResNet101 | 75.02 | [TPAMI](https://ieeexplore.ieee.org/document/9001201) [arXiv:1911.09358](https://arxiv.org/abs/1911.09358) | [Pytorch](https://github.com/MingtaoFu/gliding_vertex) | quadrilateral bbox | :white_check_mark: | | Mask OBB | ResNeXt-101 | 75.33 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/24/2930/htm) | - | attention, multi-task | :white_check_mark: | | FFA | ResNet101 | 75.7 | [ISPRS](https://www.sciencedirect.com/science/article/abs/pii/S0924271620300319) | - | enhanced FPN, rotation proposals | | | APE | ResNeXt-101(32x4) | 75.75 | [TGRS](https://ieeexplore.ieee.org/abstract/document/9057525) [arXiv:1906.09447](https://arxiv.org/abs/1906.09447) | - | adaptive period embedding, length independent IoU (LIIoU)| :white_check_mark: | | CenterMap OBB | ResNet101 | 76.03| [TGRS](https://ieeexplore.ieee.org/abstract/document/9151222) | - | center-probability-map | | | CSL | ResNet152 | 76.17 | [ECCV2020](http://arxiv.org/abs/2003.05597) | [TF:CSL_RetinaNet](https://github.com/Thinklab-SJTU/CSL_RetinaNet_Tensorflow) | angular boundary problem | :white_check_mark: | | OWSR | Ensemble (ResNet101 + ResNeXt101 + mdcn-ResNet101) | 76.36 | [CVPR2019 WorkShop](http://openaccess.thecvf.com/content_CVPRW_2019/papers/DOAI/Li_Learning_Object-Wise_Semantic_Representation_for_Detection_in_Remote_Sensing_Imagery_CVPRW_2019_paper.pdf) [TGRS](https://ieeexplore.ieee.org/abstract/document/8960460) | - | enhanced FPN | | | R³Det++ | ResNet152 | 76.56 | [arXiv:2004.13316](https://arxiv.org/abs/2004.13316) | [TF](https://github.com/SJTU-Thinklab-Det/R3Det_Tensorflow) | refined single stage, feature alignment, denoising | :white_check_mark: | | SCRDet++ | ResNet101 | 76.81 | [arXiv:2004.13316](https://arxiv.org/abs/2004.13316) | [TF](https://github.com/SJTU-Thinklab-Det/DOTA-DOAI) | angular boundary problem, denoising | :white_check_mark: | | S²A-Net | ResNet101 | 79.15 | [arXiv:2008.09397](https://arxiv.org/abs/2008.09397) | [Pytorch](https://github.com/csuhan/s2anet) | refined single stage, feature alignment | :white_check_mark: | ### DOTA1.0 (Task2) | Model | Backbone | mAP | Paper Link | Code Link | Remark | Recommend | |:-----:|:--------:|:---:|:----------:|:---------:|:---------:|:---------:| | FR-H (DOTA) | ResNet101 | 60.46 | [CVPR2018](http://openaccess.thecvf.com/content_cvpr_2018/html/Xia_DOTA_A_Large-Scale_CVPR_2018_paper.html) | [MXNet](https://github.com/jessemelpolio/Faster_RCNN_for_DOTA) | DOTA dataset, baseline | :white_check_mark: | | Deep Active Learning | ResNet18 | 64.26 | [arXiv:2003.08793](https://arxiv.org/abs/2003.08793) | - | CenterNet, Deep Active Learning | :white_check_mark: | | SBL | ResNet50 | 64.77 | [arXiv:1810.08103](https://arxiv.org/abs/1810.08103) | - | single stage | | FMSSD | VGG16 | 72.43 | [TGRS](https://ieeexplore.ieee.org/abstract/document/8930933) | - | IoU-based weighted loss, enhanced FPN | | | ICN | ResNet101 | 72.45 | [ACCV2018](https://link.springer.com/chapter/10.1007/978-3-030-20893-6_10) | - | image cascade, multi-scale | :white_check_mark: | | IoU-Adaptive R-CNN | ResNet101 | 72.72 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/3/286) | - | IoU-based weighted loss, cascade| | | EFR | VGG16 | 73.49 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/18/2095) | [Pytorch](https://github.com/pioneer2018/dtdm-di) | enhanced FPN | | | SCRDet | ResNet101 | 75.35 | [ICCV2019](http://openaccess.thecvf.com/content_ICCV_2019/papers/Yang_SCRDet_Towards_More_Robust_Detection_for_Small_Cluttered_and_Rotated_ICCV_2019_paper.pdf) | [TF](https://github.com/DetectionTeamUCAS/R2CNN-Plus-Plus_Tensorflow) | attention, angular boundary problem | :white_check_mark: | | FADet | ResNet101 | 75.38 | [ICIP2019](https://ieeexplore.ieee.org/abstract/document/8803521) | - | attention | | | MFIAR-Net | ResNet152 | 76.07 | [Sensors](https://www.mdpi.com/1424-8220/20/6/1686/htm) | - | feature attention, enhanced FPN | | | Mask OBB | ResNeXt-101 | 76.98 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/24/2930/htm) | - | attention, multi-task | :white_check_mark: | | CenterMap OBB | ResNet101 | 77.33| [TGRS](https://ieeexplore.ieee.org/abstract/document/9151222) | - | center-probability-map | | | A²RMNet | ResNet101 | 78.45 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/13/1594) | - | attention, enhanced FPN, different pooled sizes | | | OWSR | Ensemble (ResNet101 + ResNeXt101 + mdcn-ResNet101) | 78.79 | [CVPR2019 WorkShop](http://openaccess.thecvf.com/content_CVPRW_2019/papers/DOAI/Li_Learning_Object-Wise_Semantic_Representation_for_Detection_in_Remote_Sensing_Imagery_CVPRW_2019_paper.pdf) [TGRS](https://ieeexplore.ieee.org/abstract/document/8960460) | - | enhanced FPN | | | Parallel Cascade R-CNN |ResNeXt-101 | 78.96 | [Journal of Physics: Conference Series](https://iopscience.iop.org/article/10.1088/1742-6596/1544/1/012124/meta) | - | cascade rcnn | | | DM-FPN | ResNet-Based | 79.27 | [Remote Sensing](https://www.mdpi.com/2072-4292/11/7/755/) | - | enhanced FPN | | | SCRDet++ | ResNet101 | 79.35 | [arXiv:2004.13316](https://arxiv.org/abs/2004.13316) | [TF](https://github.com/SJTU-Thinklab-Det/DOTA-DOAI) | denoising | :white_check_mark: | ### DOTA1.5 (Task1) | Model | Backbone | mAP | Paper Link | Code Link | Remark | Recommend | |:-----:|:--------:|:---:|:----------:|:---------:|:---------:|:---------:| | APE | ResNeXt-101(32x4) | 78.34 | [TGRS](https://ieeexplore.ieee.org/abstract/document/9057525) [arXiv:1906.09447](https://arxiv.org/abs/1906.09447) | - | length independent IoU (LIIoU)| :white_check_mark: | | OWSR | Ensemble (ResNet101 + ResNeXt101 + mdcn-ResNet101) | 76.60 | [TGRS](https://ieeexplore.ieee.org/abstract/document/8960460) [CVPR2019 WorkShop](http://openaccess.thecvf.com/content_CVPRW_2019/papers/DOAI/Li_Learning_Object-Wise_Semantic_Representation_for_Detection_in_Remote_Sensing_Imagery_CVPRW_2019_paper.pdf) | - | enhanced FPN | | ### DOTA1.5 (Task2) | Model | Backbone | mAP | Paper Link | Code Link | Remark | Recommend | |:-----:|:--------:|:---:|:----------:|:---------:|:---------:|:---------:| | OWSR | Ensemble (ResNet101 + ResNeXt101 + mdcn-ResNet101) | 79.50 | [TGRS](https://ieeexplore.ieee.org/abstract/document/8960460) [CVPR2019 WorkShop](http://openaccess.thecvf.com/content_CVPRW_2019/papers/DOAI/Li_Learning_Object-Wise_Semantic_Representation_for_Detection_in_Remote_Sensing_Imagery_CVPRW_2019_paper.pdf) | - | enhanced FPN | | ### Related Articles | Model | Paper Link | Code Link | Remark | Recommend | |:-----:|:----------:|:---------:|:------:| :------: | | SSSDET | [ICIP2019](https://ieeexplore.ieee.org/abstract/document/8803262) [arXiv:1909.00292](https://arxiv.org/abs/1909.00292) | - | vehicle detection, lightweight | | | AVDNet | [GRSL](https://ieeexplore.ieee.org/abstract/document/8755462) [arXiv:1907.07477](https://arxiv.org/abs/1907.07477) | - | vehicle detection, small object | | | ClusDet | [ICCV2019](http://openaccess.thecvf.com/content_ICCV_2019/papers/Yang_Clustered_Object_Detection_in_Aerial_Images_ICCV_2019_paper.pdf) | [Caffe2](https://github.com/fyangneil/Clustered-Object-Detection-in-Aerial-Image) | object cluster regions | :white_check_mark: | | DMNet | [CVPR2020 WorkShop](https://arxiv.org/abs/2004.05520) | - | object cluster regions | :white_check_mark: | | OIS | [arXiv:1911.07732](https://arxiv.org/abs/1911.07732) | [related Pytorch code](https://github.com/mrlooi/rotated_maskrcnn) | Oriented Instance Segmentation | :white_check_mark: | | LR-RCNN | [arXiv:2005.14264 ](https://arxiv.org/abs/2005.14264) | - | vehicle detection | - | ## Dataset Some remote sensing related object detection dataset statistics are in [DATASET.md](DATASET.md)