自定义数据集 #214
Description
`在训练isaid数据集时，训练成功，
python tools/train.py configs/obb/oriented_rcnn_beyond/mask_rcnn_orpn_r50_fpn_1x_isaid.py
尝试使用自定义数据集cityscapes，直接在isaid相关文件上进行修改，如下：
1）configs/obb/oriented_rcnn_beyond/mask_rcnn_orpn_r50_fpn_1x_isaid.py：两处num_classes修改为8
2）configs/obb/base/datasets/isaid.py：修改数据集路径
3）mmdet/datasets/obb/isaid.py：修改class
再次运行
python tools/train.py configs/obb/oriented_rcnn_beyond/mask_rcnn_orpn_r50_fpn_1x_isaid.py
运行结果：卡在下面这个地方
2025-03-24 21:00:22,868 - mmdet - INFO - workflow: [('train', 1)], max: 12 epochs
2025-03-24 21:00:22,868 - mmdet - INFO - Checkpoints will be saved to /mnt/f/OBBDetection-master/work_dirs/mask_rcnn_orpn_r50_fpn_1x_isaid by HardDiskBackend
我的数据集标注格式：
{"file_name": "munster/munster_000173_000019_leftImg8bit.png", "height": 1024, "width": 2048, "segm_file": "munster/munster_000173_000019_gtFine_labelIds.png", "id": 499}
"categories": [{"id": 24, "name": "person"}, {"id": 25, "name": "rider"}, {"id": 26, "name": "car"}, {"id": 27, "name": "truck"}, {"id": 28, "name": "bus"}, {"id": 31, "name": "train"}, {"id": 32, "name": "motorcycle"}, {"id": 33, "name": "bicycle"}],
{"iscrowd": 0, "category_id": 24, "bbox": [949.0, 406.0, 17.0, 39.0], "area": 413, "segmentation": {"size": [1024, 2048], "counts": "l\em0:co08J6J5K4M5K10O1O2N3M2N1_Oa0N2O1O1NmbiQ1"}, "image_id": 0, "id": 0}