underwater, research记录

训练指令：

---------------+

zhangcong@win-SYS-7048GR-TR:~/data/detection/objection-detection$ ./tools/dist_train.sh configs/underwater/cas_r50/cascade_rcnn_r50_fpn_1x.py 2 --validate

~/data/mmdetection$ ./tools/dist_train.sh configs/underwater/cascade_rcnn_r50_fpn_1x.py 2 --validate

validate表示 每次进行验证

报错：

File "/data/zhangcong/mmdetection/mmdet/datasets/custom.py", line 186, in evaluate

    raise KeyError('metric {} is not supported'.format(metric))

配置文件中加了：evaluation = dict(interval=1, metric='bbox')


编辑配置文件：

~/data/mmdetection$ vim configs/underwater/cascade_rcnn_r50_fpn_1x.py


实验2：

为了加快速度检测代码，训练和验证都用验证集,所有在log中

map很高，s0.acc和s1.acc高，s2.acc不高

配置文件中的evaluation = dict(interval=1, metric='mAP') 配合指令中 --validate 正常。



改为evaluation = dict(interval=1, metric='bbox')则 报错：

Traceback (most recent call last):

  File "./tools/train.py", line 151, in <module>

    main()

  File "./tools/train.py", line 147, in main

    meta=meta)

  File "/data/zhangcong/mmdetection/mmdet/apis/train.py", line 165, in train_detector

    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 380, in run

    epoch_runner(data_loaders[i], **kwargs)

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 288, in train

    self.call_hook('after_train_epoch')

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 241, in call_hook

    getattr(hook, fn_name)(self)

  File "/data/zhangcong/mmdetection/mmdet/core/evaluation/eval_hooks.py", line 76, in after_train_epoch

    self.evaluate(runner, results)

  File "/data/zhangcong/mmdetection/mmdet/core/evaluation/eval_hooks.py", line 33, in evaluate

    results, logger=runner.logger, **self.eval_kwargs)

  File "/data/zhangcong/mmdetection/mmdet/datasets/custom.py", line 186, in evaluate

    raise KeyError('metric {} is not supported'.format(metric))

KeyError: 'metric bbox is not supported'

Traceback (most recent call last):

  File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main

    "__main__", mod_spec)

  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code

    exec(code, run_globals)

  File "/usr/local/lib/python3.5/dist-packages/torch/distributed/launch.py", line 235, in <module>

    main()

  File "/usr/local/lib/python3.5/dist-packages/torch/distributed/launch.py", line 231, in main

    cmd=process.args)

subprocess.CalledProcessError: Command '['/usr/bin/python3', '-u', './tools/train.py', '--local_rank=0', 'configs/underwater/cascade_rcnn_r50_fpn_1x.py', '--launcher', 'pytorch', '--validate']' returned non-zero exit status 1

实验3：

训练和验证集正常划分训练



dataset中的underwater.py中增加了coco.py里的内容，包括eval, 测试指令：

python3 tools/test.py configs/underwater/cascade_rcnn_r50_fpn_1x.py work_dirs/cascade_rcnn_r50_fpn_1x/cas_r50/epoch_15shiyan3.pth --out ./result/result_epoch_15shiyan3.pkl --eval bbox

通过：

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.252

Average Precision  (AP) @[ IoU=0.50      | area=  all | maxDets=100 ] = 0.444

Average Precision  (AP) @[ IoU=0.75      | area=  all | maxDets=100 ] = 0.261

Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.140

Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.248

Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.280

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets=  1 ] = 0.097

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets= 10 ] = 0.282

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets=100 ] = 0.299

Average Recall    (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.236

Average Recall    (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.299

Average Recall    (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.317



5] lr: 0.00500, eta: 7:10:58, time: 0.785, data_time: 0.009, memory: 9044, loss_rpn_bbox: 0.0344, s2.loss_cls: 0.0356, s1.loss_cls: 0.0868, loss_rpn_cls: 0.1759, s0.acc: 94.2129, s0.loss_bbox: 0.0977, s1.loss_bbox: 0.0626, s2.acc: 96.2590, s1.acc: 95.3369, s0.loss_cls: 0.2210, s2.loss_bbox: 0.0238, loss: 0.7379

2020-05-14 03:32:14,742 - mmdet - INFO - Epoch [1][2200/2335]  lr: 0.00500, eta: 7:10:21, time: 0.790, data_time: 0.010, memory: 9044, loss_rpn_bbox: 0.0402, s2.loss_cls: 0.0373, s1.loss_cls: 0.0975, loss_rpn_cls: 0.1723, s0.acc: 93.1465, s0.loss_bbox: 0.1315, s1.loss_bbox: 0.0873, s2.acc: 96.0391, s1.acc: 94.4727, s0.loss_cls: 0.2558, s2.loss_bbox: 0.0262, loss: 0.8482

2020-05-14 03:32:52,139 - mmdet - INFO - Epoch [1][2250/2335]  lr: 0.00500, eta: 7:09:14, time: 0.748, data_time: 0.009, memory: 9044, loss_rpn_bbox: 0.0318, s2.loss_cls: 0.0323, s1.loss_cls: 0.0785, loss_rpn_cls: 0.1807, s0.acc: 94.9355, s0.loss_bbox: 0.0879, s1.loss_bbox: 0.0654, s2.acc: 96.4941, s1.acc: 95.7637, s0.loss_cls: 0.1947, s2.loss_bbox: 0.0223, loss: 0.6936

2020-05-14 03:33:31,941 - mmdet - INFO - Epoch [1][2300/2335]  lr: 0.00500, eta: 7:08:42, time: 0.796, data_time: 0.010, memory: 9044, loss_rpn_bbox: 0.0304, s2.loss_cls: 0.0321, s1.loss_cls: 0.0847, loss_rpn_cls: 0.1587, s0.acc: 94.1387, s0.loss_bbox: 0.1017, s1.loss_bbox: 0.0707, s2.acc: 96.5566, s1.acc: 95.1637, s0.loss_cls: 0.2136, s2.loss_bbox: 0.0227, loss: 0.7147

[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 800/800, 6.9 task/s, elapsed: 115s, ETA:    0s

Traceback (most recent call last):

  File "./tools/train.py", line 151, in <module>

    main()

  File "./tools/train.py", line 147, in main

    meta=meta)

  File "/data/zhangcong/mmdetection/mmdet/apis/train.py", line 165, in train_detector

    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 380, in run

    epoch_runner(data_loaders[i], **kwargs)

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 288, in train

    self.call_hook('after_train_epoch')

  File "/usr/local/lib/python3.5/dist-packages/mmcv/runner/runner.py", line 241, in call_hook

    getattr(hook, fn_name)(self)

  File "/data/zhangcong/mmdetection/mmdet/core/evaluation/eval_hooks.py", line 76, in after_train_epoch

    self.evaluate(runner, results)

  File "/data/zhangcong/mmdetection/mmdet/core/evaluation/eval_hooks.py", line 33, in evaluate

    results, logger=runner.logger, **self.eval_kwargs)

  File "/data/zhangcong/mmdetection/mmdet/datasets/underwater.py", line 311, in evaluate

    raise KeyError('metric {} is not supported'.format(metric))

KeyError: 'metric mAP is not supported'

Traceback (most recent call last):

  File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main

    "__main__", mod_spec)

  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code

    exec(code, run_globals)

  File "/usr/local/lib/python3.5/dist-packages/torch/distributed/launch.py", line 235, in <module>

    main()

  File "/usr/local/lib/python3.5/dist-packages/torch/distributed/launch.py", line 231, in main

    cmd=process.args)

subprocess.CalledProcessError: Command '['/usr/bin/python3', '-u', './tools/train.py', '--local_rank=0', 'configs/underwater/cascade_rcnn_r50_fpn_1x.py', '--launcher', 'pytorch', '--validate']' returned non-zero exit status 1



python3 tools/test.py configs/underwater/cascade_rcnn_r50_fpn_1x.py work_dirs/cascade_rcnn_r50_fpn_1x/epoch_24.pth --out ./result/result_epoch_24.pkl --eval bbox --show



python3 setup.py develop

sudo python3 setup.py install


python3 ./tools/train.py configs/underwater/cas_r50/cascade_rcnn_r50_fpn_1x.py 2 --validate




zhangcong@win-SYS-7048GR-TR:~/data/detection/objection-detection$ python3 tools/test.py configs/underwater/cas_r50/cascade_rcnn_r50_fpn_1x.py work_dirs/cascade_rcnn_r50_fpn_1x/epoch_2.pth --out ./result/result_epoch_2.pkl --eval bbox

loading annotations into memory...

Done (t=0.03s)

creating index...

index created!

[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 800/800, 1.3 task/s, elapsed: 613s, ETA:    0s

writing results to ./result/result_epoch_2.pkl

Starting evaluate bbox

Loading and preparing results...

DONE (t=0.59s)

creating index...

index created!

Running per image evaluation...

Evaluate annotation type *bbox*

DONE (t=7.90s).

Accumulating evaluation results...

DONE (t=0.98s).

Average Precision  (AP) @[ IoU=0.50:0.95 | area=  all | maxDets=100 ] = 0.365

Average Precision  (AP) @[ IoU=0.50      | area=  all | maxDets=100 ] = 0.685

Average Precision  (AP) @[ IoU=0.75      | area=  all | maxDets=100 ] = 0.341

Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.147

Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.336

Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.409

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets=  1 ] = 0.154

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets= 10 ] = 0.465

Average Recall    (AR) @[ IoU=0.50:0.95 | area=  all | maxDets=100 ] = 0.559

Average Recall    (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.374

Average Recall    (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.546

Average Recall    (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.580

最后编辑于 ：
©著作权归作者所有,转载或内容合作请联系作者
  • 人面猴
    序言：七十年代末，一起剥皮案震惊了整个滨河市，随后出现的几起案子，更是在滨河造成了极大的恐慌，老刑警刘岩，带你破解...
    沈念sama阅读 194,088评论 5赞 459
  • 死咒
    序言：滨河连续发生了三起死亡事件，死亡现场离奇诡异，居然都是意外死亡，警方通过查阅死者的电脑和手机，发现死者居然都...
    沈念sama阅读 81,715评论 2赞 371
  • 救了他两次的神仙让他今天三更去死
    文/潘晓璐 我一进店门，熙熙楼的掌柜王于贵愁眉苦脸地迎上来，“玉大人，你说我怎么就摊上这事。” “怎么了？”我有些...
    开封第一讲书人阅读 141,361评论 0赞 319
  • 道士缉凶录：失踪的卖姜人
    文/不坏的土叔 我叫张陵，是天一观的道长。 经常有香客问我，道长，这世上最难降的妖魔是什么？ 我笑而不...
    开封第一讲书人阅读 52,099评论 1赞 263
  • 港岛之恋（遗憾婚礼）
    正文 为了忘掉前任，我火速办了婚礼，结果婚礼上，老公的妹妹穿的比我还像新娘。我一直安慰自己，他们只是感情好，可当我...
    茶点故事阅读 60,987评论 4赞 355
  • 恶毒庶女顶嫁案：这布局不是一般人想出来的
    文/花漫 我一把揭开白布。 她就那样静静地躺着，像睡着了一般。 火红的嫁衣衬着肌肤如雪。 梳的纹丝不乱的头发上，一...
    开封第一讲书人阅读 46,063评论 1赞 272
  • 城市分裂传说
    那天，我揣着相机与录音，去河边找鬼。 笑死，一个胖子当着我的面吹牛，可吹牛的内容都是我干的。 我是一名探鬼主播，决...
    沈念sama阅读 36,486评论 3赞 381
  • 双鸳鸯连环套：你想象不到人心有多黑
    文/苍兰香墨 我猛地睁开眼，长吁一口气：“原来是场噩梦啊……” “哼！你这毒妇竟也来了？” 一声冷哼从身侧响起，我...
    开封第一讲书人阅读 35,175评论 0赞 253
  • 万荣杀人案实录
    序言：老挝万荣一对情侣失踪，失踪者是张志新（化名）和其女友刘颖，没想到半个月后，有当地人在树林里发现了一具尸体，经...
    沈念sama阅读 39,440评论 1赞 290
  • 护林员之死
    正文 独居荒郊野岭守林人离奇死亡，尸身上长有42处带血的脓包…… 初始之章·张勋 以下内容为张勋视角 年9月15日...
    茶点故事阅读 34,518评论 2赞 309
  • 白月光启示录
    正文 我和宋清朗相恋三年，在试婚纱的时候发现自己被绿了。 大学时的朋友给我发了我未婚夫和他白月光在一起吃饭的照片。...
    茶点故事阅读 36,305评论 1赞 326
  • 活死人
    序言：一个原本活蹦乱跳的男人离奇死亡，死状恐怖，灵堂内的尸体忽然破棺而出，到底是诈尸还是另有隐情，我是刑警宁泽，带...
    沈念sama阅读 32,190评论 3赞 312
  • 日本核电站爆炸内幕
    正文 年R本政府宣布，位于F岛的核电站，受9级特大地震影响，放射性物质发生泄漏。R本人自食恶果不足惜，却给世界环境...
    茶点故事阅读 37,550评论 3赞 298
  • 男人毒药：我在死后第九天来索命
    文/蒙蒙 一、第九天 我趴在偏房一处隐蔽的房顶上张望。 院中可真热闹，春花似锦、人声如沸。这庄子的主人今日做“春日...
    开封第一讲书人阅读 28,880评论 0赞 17
  • 一桩弑父案，背后竟有这般阴谋
    文/苍兰香墨 我抬头看了看天上的太阳。三九已至，却和暖如春，着一层夹袄步出监牢的瞬间，已是汗流浃背。 一阵脚步声响...
    开封第一讲书人阅读 30,152评论 1赞 250
  • 情欲美人皮
    我被黑心中介骗来泰国打工， 没想到刚下飞机就差点儿被人妖公主榨干…… 1. 我叫王不留，地道东北人。 一个月前我还...
    沈念sama阅读 41,451评论 2赞 341
  • 代替公主和亲
    正文 我出身青楼，却偏偏与公主长得像，于是被迫代替她去往敌国和亲。 传闻我的和亲对象是个残疾皇子，可洞房花烛夜当晚...
    茶点故事阅读 40,637评论 2赞 335