AerialDetection
AerialDetection copied to clipboard
dingjiansw101
Hi， 作者您好。我用该框架的[faster_rcnn_r50_fpn_1x_dota.py](https://github.com/dingjiansw101/AerialDetection/blob/master/configs/DOTA/faster_rcnn_r50_fpn_1x_dota.py)跑了DOTA V1.0的task2， 除了data文件，其余的配置比如模型、优化器等完全保持不变，没有做任何修改。下面是我的data config： ```python # train trainval = [] train = dict( type=dataset_type, ann_file=data_root + 'official_v10_train1024/train1024_v10.json', img_prefix=data_root + 'official_v10_train1024/images/', img_scale=(1024, 1024), img_norm_cfg=img_norm_cfg, size_divisor=32, flip_ratio=0.5, with_mask=True, with_crowd=True, with_label=True) trainval.append(train)...
博主，您好，DOTA数据集中的测试集里面只有图片，没有标注文件.txt，这样的test能用于mmdetection的test.py测试吗？我看见论文里面好多都有dota数据集测试的结果，他们用的是测试集吗？还是验证集？为什么dota数据集的测试集没有标注文件官网还提供了生成测试集的json文件？生产的测试集的json文件真的有用吗？
i upload the nms resaults and get this: mAP: 0.3745174540144275 ap of each class: plane:0.5300541025955471, baseball-diamond:0.27181818181818185, bridge:0.43130126717635453, ground-track-field:0.14171122994652408, small-vehicle:0.24575509050331162, large-vehicle:0.25701370866991136, ship:0.7015457849808598, tennis-court:0.5453346126169345, basketball-court:0.3152847152847153, storage-tank:0.09090909090909091, soccer-ball-field:0.18932806324110674, roundabout:0.1655011655011655, harbor:0.6844767188033766, swimming-pool:0.6143607552420431, helicopter:0.4333673229272884 why...
### I run the compile.sh with this step Compile cuda extensions. ./compile.sh ### but it has some problem as following: /media/miki/ubuntu/deep_learning/huoyan/AerialDetection-master/mmdet/ops/poly_nms/src/poly_nms_cuda.cpp:3:23: error: ‘AT_CHECK’ was not declared in this scope #define...
I just want to ask, have you tried to implement the AerialDetection in the latest version of MMDetection?
Hi, I am getting an error on running the following lines on google colab. May be someone else faced this issue before. 1. !python tools/test.py configs/DOTA/faster_rcnn_RoITrans_r50_fpn_1x_dota.py work_dirs/faster_rcnn_RoITrans_r50_fpn_1x_dota/epoch_12.pth --out work_dirs/faster_rcnn_RoITrans_r50_fpn_1x_dota/results.pkl 2....
您好，非常感谢您的工作。我想问一些关于数据增强方面的问题。 1.我有注意到你在faster_rcnn_RoITrans_r50_fpn_1x_dota1_5_gap512_msTrainTest_rotationTrainTest.py 这个文件中提到了用ms(muti-scale)多尺度的训练方式， 但是在train_config中的img_scale=(1024, 1024)只有一个尺度，请问你是在将数据split的时候，用了多尺度的方式吗？ 2.为什么不使用mmdetection中 img_scale=[(1024, 1024),(608, 608),(816,816)] 这种在网络训练过程中随机选取一个尺度的形式呢？ 3.在test.py函数中 是直接输入整张图进行预测的，如果test图片比较大，要resize到（1024，1024）会丢失很多信息，为什么在test函数中没有采用demo_large_image.py中将大图crop成小图的形式呢？ 期待您的回复！
how can i get more evaluating without DOTA server(its just give AP):i need to get more like recall precision curve or APs APm ,... after training it gives log.json file...
