SwinTextSpotter batch

REC_HEAD: BATCH_SIZE: 128 SOLVER: IMS_PER_BATCH: 8 IMS_PER_BATCH是什么

Sep 11 '23 13:09 Huangqqqhhh

IMS_PER_BATCH就是每个batch里面图片的数量，REC_HEAD下面的BATCH_SIZE是识别的BATCH_SIZE

Sep 22 '23 02:09 mxin262

非常感谢您的回复，我还有一些问题： 1）我的设备是2张3090，在训练中文数据集的时候，IMS_PER_BATCH要调到2（英文时是4），这是因为中文数据集图片更大嘛？

2）在预训练中文数据集的时候，大概每隔一段时间，8万次迭代左右吧，就会出现一个报错，batch size RuntimeError: The size of tensor a (302) must match the size of tensor b (323) at non-singleton dimension 0 后面调整识别头BATCH_SIZE之后，从8万次迭代继续训练，然后在16万、24万多次迭代之后又出现了新的错误，RuntimeError: cannot reshape tensor of 0 elements into shape [0, 256, -1] because the unspecified dimension size -1 can be any value and is ambiguous，为什么每隔8万多次之后会出现错误，而后还可以继续训练？

3）在测试IC15数据集的时候，我发现召回率、准确率等为0，是因为它的注释文件不对嘛？

希望您能给我解答一下，谢谢！

权是海 @.***

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年9月22日(星期五) 上午10:07 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [mxin262/SwinTextSpotter] batch_size (Issue #102)

IMS_PER_BATCH就是每个batch里面图片的数量，REC_HEAD下面的BATCH_SIZE是识别的BATCH_SIZE

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

Sep 22 '23 02:09 Huangqqqhhh

1）中文的类别更多，识别分辨率更大了，所以占用显存更多； 2）这是因为最大的文本数量限制了只能300个，调整识别头BATCH_SIZE之后，出现这个报错是因为图片中没有文本实例，可以继续训练的， 3）可以试试把文件先下载好，根据readme来评测

Sep 22 '23 03:09 mxin262

非常感谢您的回复对于第2个问题，IMS_PER_BATCH和REC_HEAD下面的BATCH_SIZE存在比例关系嘛，因为我调整IMS_PER_BATCH为2这个时候，不知道后面识别的BATCH_SIZE应该设置为多少

权是海 @.***

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年9月22日(星期五) 中午11:03 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [mxin262/SwinTextSpotter] batch_size (Issue #102)

1）中文的类别更多，识别分辨率更大了，所以占用显存更多； 2）这是因为最大的文本数量限制了只能300个，调整识别头BATCH_SIZE之后，出现这个报错是因为图片中没有文本实例，可以继续训练的， 3）可以试试把文件先下载好，根据readme来评测

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

Sep 22 '23 03:09 Huangqqqhhh

根据你的显存来设置就好了

Sep 22 '23 03:09 mxin262

[10/19 14:04:32 d2.engine.hooks]: Overall training speed: 19998 iterations in 3:15:31 (0.5866 s / it) [10/19 14:04:32 d2.engine.hooks]: Total training time: 3:16:46 (0:01:14 on hooks) [10/19 14:04:32 d2.data.datasets.coco]: Loaded 300 images in COCO format from datasets/totaltext/totaltext_test.json [10/19 14:04:32 d2.data.build]: Distribution of instances among all 1 categories:

category	#instances
text	1975

[10/19 14:04:32 d2.data.dataset_mapper]: [DatasetMapper] Augmentations used in inference: [ResizeShortestEdge(short_edge_length=(1000, 1000), max_size=1824, sample_style='choice')]
[10/19 14:04:32 d2.data.common]: Serializing 300 elements to byte tensors and concatenating them all ...
[10/19 14:04:32 d2.data.common]: Serialized dataset takes 0.54 MiB
[10/19 14:04:32 d2.evaluation.evaluator]: Start inference on 150 images
/home/hhq/data/anaconda3/envs/SWINTS/lib/python3.8/site-packages/shapely/set_operations.py:133: RuntimeWarning: invalid value encountered in intersection
return lib.intersection(a, b, **kwargs)
/home/hhq/data/anaconda3/envs/SWINTS/lib/python3.8/site-packages/shapely/set_operations.py:133: RuntimeWarning: invalid value encountered in intersection
return lib.intersection(a, b, **kwargs)
[10/19 14:04:50 d2.evaluation.evaluator]: Inference done 1/150. 6.7070 s / img. ETA=0:44:16
[10/19 14:04:59 d2.evaluation.evaluator]: Inference done 2/150. 6.6461 s / img. ETA=0:32:28
[10/19 14:05:08 d2.evaluation.evaluator]: Inference done 3/150. 6.6834 s / img. ETA=0:28:55
[10/19 14:05:16 d2.evaluation.evaluator]: Inference done 4/150. 6.6690 s / img. ETA=0:26:43
[10/19 14:05:25 d2.evaluation.evaluator]: Inference done 5/150. 6.6684 s / img. ETA=0:25:26
[10/19 14:05:33 d2.evaluation.evaluator]: Inference done 6/150. 6.6617 s / img. ETA=0:19:18
[10/19 14:05:41 d2.evaluation.evaluator]: Inference done 7/150. 6.6557 s / img. ETA=0:19:08
[10/19 14:05:52 d2.evaluation.evaluator]: Inference done 8/150. 6.6652 s / img. ETA=0:21:28
[10/19 14:06:01 d2.evaluation.evaluator]: Inference done 9/150. 6.6563 s / img. ETA=0:20:51
[10/19 14:06:09 d2.evaluation.evaluator]: Inference done 10/150. 6.6547 s / img. ETA=0:20:26
[10/19 14:06:17 d2.evaluation.evaluator]: Inference done 11/150. 6.6557 s / img. ETA=0:19:58
[10/19 14:06:25 d2.evaluation.evaluator]: Inference done 12/150. 6.6504 s / img. ETA=0:19:43
[10/19 14:06:33 d2.evaluation.evaluator]: Inference done 13/150. 6.6458 s / img. ETA=0:19:26
[10/19 14:06:42 d2.evaluation.evaluator]: Inference done 14/150. 6.6475 s / img. ETA=0:19:27
[10/19 14:06:52 d2.evaluation.evaluator]: Inference done 15/150. 6.6845 s / img. ETA=0:19:30
[10/19 14:07:01 d2.evaluation.evaluator]: Inference done 16/150. 6.6815 s / img. ETA=0:19:30
[10/19 14:07:11 d2.evaluation.evaluator]: Inference done 17/150. 6.6790 s / img. ETA=0:19:31
[10/19 14:07:20 d2.evaluation.evaluator]: Inference done 18/150. 6.6772 s / img. ETA=0:19:24
[10/19 14:07:28 d2.evaluation.evaluator]: Inference done 19/150. 6.6755 s / img. ETA=0:19:10
[10/19 14:07:37 d2.evaluation.evaluator]: Inference done 20/150. 6.6706 s / img. ETA=0:18:58
[10/19 14:07:46 d2.evaluation.evaluator]: Inference done 21/150. 6.6681 s / img. ETA=0:18:53
[10/19 14:07:55 d2.evaluation.evaluator]: Inference done 22/150. 6.6669 s / img. ETA=0:18:45
[10/19 14:08:05 d2.evaluation.evaluator]: Inference done 23/150. 6.6686 s / img. ETA=0:18:49
[10/19 14:08:14 d2.evaluation.evaluator]: Inference done 24/150. 6.6662 s / img. ETA=0:18:41
[10/19 14:08:22 d2.evaluation.evaluator]: Inference done 25/150. 6.6650 s / img. ETA=0:18:25
[10/19 14:08:30 d2.evaluation.evaluator]: Inference done 26/150. 6.6614 s / img. ETA=0:18:13
[10/19 14:08:39 d2.evaluation.evaluator]: Inference done 27/150. 6.6599 s / img. ETA=0:18:02
[10/19 14:08:47 d2.evaluation.evaluator]: Inference done 28/150. 6.6586 s / img. ETA=0:17:51
[10/19 14:08:55 d2.evaluation.evaluator]: Inference done 29/150. 6.6547 s / img. ETA=0:17:38
[10/19 14:09:04 d2.evaluation.evaluator]: Inference done 30/150. 6.6556 s / img. ETA=0:17:31
[10/19 14:09:12 d2.evaluation.evaluator]: Inference done 31/150. 6.6533 s / img. ETA=0:17:19
[10/19 14:09:21 d2.evaluation.evaluator]: Inference done 32/150. 6.6937 s / img. ETA=0:17:12
[10/19 14:09:32 d2.evaluation.evaluator]: Inference done 33/150. 6.6920 s / img. ETA=0:17:09
[10/19 14:09:41 d2.evaluation.evaluator]: Inference done 34/150. 6.6887 s / img. ETA=0:17:04
[10/19 14:09:50 d2.evaluation.evaluator]: Inference done 35/150. 6.6866 s / img. ETA=0:16:56
[10/19 14:10:01 d2.evaluation.evaluator]: Inference done 36/150. 6.6862 s / img. ETA=0:16:54
[10/19 14:10:09 d2.evaluation.evaluator]: Inference done 37/150. 6.6865 s / img. ETA=0:16:43
[10/19 14:10:17 d2.evaluation.evaluator]: Inference done 38/150. 6.6852 s / img. ETA=0:16:32
[10/19 14:10:25 d2.evaluation.evaluator]: Inference done 39/150. 6.6857 s / img. ETA=0:16:20
[10/19 14:10:34 d2.evaluation.evaluator]: Inference done 40/150. 6.6851 s / img. ETA=0:16:10
[10/19 14:10:42 d2.evaluation.evaluator]: Inference done 41/150. 6.6838 s / img. ETA=0:16:00
[10/19 14:10:51 d2.evaluation.evaluator]: Inference done 42/150. 6.6828 s / img. ETA=0:15:52
[10/19 14:10:59 d2.evaluation.evaluator]: Inference done 43/150. 6.6822 s / img. ETA=0:15:41
[10/19 14:11:08 d2.evaluation.evaluator]: Inference done 44/150. 6.6828 s / img. ETA=0:15:30
[10/19 14:11:18 d2.evaluation.evaluator]: Inference done 45/150. 6.6826 s / img. ETA=0:15:25
[10/19 14:11:27 d2.evaluation.evaluator]: Inference done 46/150. 6.6824 s / img. ETA=0:15:17
[10/19 14:11:35 d2.evaluation.evaluator]: Inference done 47/150. 6.6813 s / img. ETA=0:15:07
[10/19 14:11:44 d2.evaluation.evaluator]: Inference done 48/150. 6.6809 s / img. ETA=0:14:59
[10/19 14:11:53 d2.evaluation.evaluator]: Inference done 49/150. 6.6805 s / img. ETA=0:14:49
/home/hhq/data/anaconda3/envs/SWINTS/lib/python3.8/site-packages/shapely/set_operations.py:133: RuntimeWarning: divide by zero encountered in intersection
return lib.intersection(a, b, **kwargs)
[10/19 14:12:03 d2.evaluation.evaluator]: Inference done 50/150. 6.6798 s / img. ETA=0:14:43
[10/19 14:12:12 d2.evaluation.evaluator]: Inference done 51/150. 6.6795 s / img. ETA=0:14:35
[10/19 14:12:20 d2.evaluation.evaluator]: Inference done 52/150. 6.6785 s / img. ETA=0:14:24
[10/19 14:12:28 d2.evaluation.evaluator]: Inference done 53/150. 6.6778 s / img. ETA=0:14:15
[10/19 14:12:37 d2.evaluation.evaluator]: Inference done 54/150. 6.6787 s / img. ETA=0:14:06
[10/19 14:12:45 d2.evaluation.evaluator]: Inference done 55/150. 6.6810 s / img. ETA=0:13:56
[10/19 14:12:55 d2.evaluation.evaluator]: Inference done 56/150. 6.6803 s / img. ETA=0:13:48
[10/19 14:13:03 d2.evaluation.evaluator]: Inference done 57/150. 6.6797 s / img. ETA=0:13:38
[10/19 14:13:10 d2.evaluation.evaluator]: Inference done 58/150. 6.6794 s / img. ETA=0:13:27
[10/19 14:13:20 d2.evaluation.evaluator]: Inference done 59/150. 6.6784 s / img. ETA=0:13:20
[10/19 14:13:28 d2.evaluation.evaluator]: Inference done 60/150. 6.6782 s / img. ETA=0:13:10
[10/19 14:13:42 d2.evaluation.evaluator]: Inference done 61/150. 6.6795 s / img. ETA=0:13:08
[10/19 14:13:50 d2.evaluation.evaluator]: Inference done 62/150. 6.6785 s / img. ETA=0:12:59
[10/19 14:13:58 d2.evaluation.evaluator]: Inference done 63/150. 6.6792 s / img. ETA=0:12:49
[10/19 14:14:07 d2.evaluation.evaluator]: Inference done 64/150. 6.6780 s / img. ETA=0:12:40
[10/19 14:14:16 d2.evaluation.evaluator]: Inference done 65/150. 6.6773 s / img. ETA=0:12:31
[10/19 14:14:29 d2.evaluation.evaluator]: Inference done 66/150. 6.6766 s / img. ETA=0:12:28
[10/19 14:14:37 d2.evaluation.evaluator]: Inference done 67/150. 6.6761 s / img. ETA=0:12:18
[10/19 14:14:48 d2.evaluation.evaluator]: Inference done 68/150. 6.6757 s / img. ETA=0:12:12
[10/19 14:14:56 d2.evaluation.evaluator]: Inference done 69/150. 6.6754 s / img. ETA=0:12:02
[10/19 14:15:06 d2.evaluation.evaluator]: Inference done 70/150. 6.6753 s / img. ETA=0:11:55
[10/19 14:15:20 d2.evaluation.evaluator]: Inference done 71/150. 6.6743 s / img. ETA=0:11:52
[10/19 14:15:30 d2.evaluation.evaluator]: Inference done 72/150. 6.6746 s / img. ETA=0:11:44
[10/19 14:15:38 d2.evaluation.evaluator]: Inference done 73/150. 6.6747 s / img. ETA=0:11:34
[10/19 14:15:50 d2.evaluation.evaluator]: Inference done 74/150. 6.6750 s / img. ETA=0:11:28
[10/19 14:16:02 d2.evaluation.evaluator]: Inference done 75/150. 6.6766 s / img. ETA=0:11:22
[10/19 14:16:15 d2.evaluation.evaluator]: Inference done 76/150. 6.6777 s / img. ETA=0:11:17
[10/19 14:16:24 d2.evaluation.evaluator]: Inference done 77/150. 6.6783 s / img. ETA=0:11:08
[10/19 14:16:34 d2.evaluation.evaluator]: Inference done 78/150. 6.6790 s / img. ETA=0:10:59
[10/19 14:16:44 d2.evaluation.evaluator]: Inference done 79/150. 6.6809 s / img. ETA=0:10:51
[10/19 14:16:54 d2.evaluation.evaluator]: Inference done 80/150. 6.6841 s / img. ETA=0:10:42
[10/19 14:17:06 d2.evaluation.evaluator]: Inference done 81/150. 6.6880 s / img. ETA=0:10:36
[10/19 14:17:16 d2.evaluation.evaluator]: Inference done 82/150. 6.6935 s / img. ETA=0:10:27
[10/19 14:17:24 d2.evaluation.evaluator]: Inference done 83/150. 6.6935 s / img. ETA=0:10:17
[10/19 14:17:34 d2.evaluation.evaluator]: Inference done 84/150. 6.6946 s / img. ETA=0:10:08
[10/19 14:17:42 d2.evaluation.evaluator]: Inference done 85/150. 6.6952 s / img. ETA=0:09:58
[10/19 14:17:52 d2.evaluation.evaluator]: Inference done 86/150. 6.6951 s / img. ETA=0:09:50
[10/19 14:18:01 d2.evaluation.evaluator]: Inference done 87/150. 6.6942 s / img. ETA=0:09:40
[10/19 14:18:09 d2.evaluation.evaluator]: Inference done 88/150. 6.6936 s / img. ETA=0:09:30
[10/19 14:18:19 d2.evaluation.evaluator]: Inference done 89/150. 6.6924 s / img. ETA=0:09:22
[10/19 14:18:28 d2.evaluation.evaluator]: Inference done 90/150. 6.6920 s / img. ETA=0:09:12
[10/19 14:18:37 d2.evaluation.evaluator]: Inference done 91/150. 6.6910 s / img. ETA=0:09:02
[10/19 14:18:49 d2.evaluation.evaluator]: Inference done 92/150. 6.6902 s / img. ETA=0:08:55
[10/19 14:18:57 d2.evaluation.evaluator]: Inference done 93/150. 6.6901 s / img. ETA=0:08:45
[10/19 14:19:11 d2.evaluation.evaluator]: Inference done 94/150. 6.6905 s / img. ETA=0:08:39
[10/19 14:19:19 d2.evaluation.evaluator]: Inference done 95/150. 6.6904 s / img. ETA=0:08:29
[10/19 14:19:28 d2.evaluation.evaluator]: Inference done 96/150. 6.6899 s / img. ETA=0:08:20
[10/19 14:19:43 d2.evaluation.evaluator]: Inference done 97/150. 6.6900 s / img. ETA=0:08:14
[10/19 14:19:53 d2.evaluation.evaluator]: Inference done 98/150. 6.6965 s / img. ETA=0:08:05
[10/19 14:20:01 d2.evaluation.evaluator]: Inference done 99/150. 6.6970 s / img. ETA=0:07:55
[10/19 14:20:09 d2.evaluation.evaluator]: Inference done 100/150. 6.6964 s / img. ETA=0:07:45
[10/19 14:20:30 d2.evaluation.evaluator]: Inference done 101/150. 6.6980 s / img. ETA=0:07:41
[10/19 14:20:39 d2.evaluation.evaluator]: Inference done 102/150. 6.6979 s / img. ETA=0:07:32
[10/19 14:20:51 d2.evaluation.evaluator]: Inference done 103/150. 6.6984 s / img. ETA=0:07:24
[10/19 14:21:03 d2.evaluation.evaluator]: Inference done 104/150. 6.6989 s / img. ETA=0:07:15
[10/19 14:21:13 d2.evaluation.evaluator]: Inference done 105/150. 6.6995 s / img. ETA=0:07:06
[10/19 14:21:25 d2.evaluation.evaluator]: Inference done 106/150. 6.7007 s / img. ETA=0:06:58
[10/19 14:21:38 d2.evaluation.evaluator]: Inference done 107/150. 6.7005 s / img. ETA=0:06:50
[10/19 14:21:46 d2.evaluation.evaluator]: Inference done 108/150. 6.7015 s / img. ETA=0:06:40
[10/19 14:22:01 d2.evaluation.evaluator]: Inference done 109/150. 6.7015 s / img. ETA=0:06:32
[10/19 14:22:09 d2.evaluation.evaluator]: Inference done 110/150. 6.7014 s / img. ETA=0:06:22
[10/19 14:22:21 d2.evaluation.evaluator]: Inference done 111/150. 6.7029 s / img. ETA=0:06:13
[10/19 14:22:30 d2.evaluation.evaluator]: Inference done 112/150. 6.7051 s / img. ETA=0:06:03
[10/19 14:22:38 d2.evaluation.evaluator]: Inference done 113/150. 6.7058 s / img. ETA=0:05:53
[10/19 14:22:49 d2.evaluation.evaluator]: Inference done 114/150. 6.7059 s / img. ETA=0:05:44
[10/19 14:22:58 d2.evaluation.evaluator]: Inference done 115/150. 6.7059 s / img. ETA=0:05:34
[10/19 14:23:06 d2.evaluation.evaluator]: Inference done 116/150. 6.7063 s / img. ETA=0:05:25
[10/19 14:23:14 d2.evaluation.evaluator]: Inference done 117/150. 6.7068 s / img. ETA=0:05:15
[10/19 14:23:24 d2.evaluation.evaluator]: Inference done 118/150. 6.7083 s / img. ETA=0:05:05
[10/19 14:23:35 d2.evaluation.evaluator]: Inference done 119/150. 6.7083 s / img. ETA=0:04:56
[10/19 14:23:44 d2.evaluation.evaluator]: Inference done 120/150. 6.7079 s / img. ETA=0:04:46
[10/19 14:23:55 d2.evaluation.evaluator]: Inference done 121/150. 6.7080 s / img. ETA=0:04:37
[10/19 14:24:04 d2.evaluation.evaluator]: Inference done 122/150. 6.7075 s / img. ETA=0:04:27
[10/19 14:24:13 d2.evaluation.evaluator]: Inference done 123/150. 6.7069 s / img. ETA=0:04:18
[10/19 14:24:21 d2.evaluation.evaluator]: Inference done 124/150. 6.7061 s / img. ETA=0:04:08
[10/19 14:24:29 d2.evaluation.evaluator]: Inference done 125/150. 6.7052 s / img. ETA=0:03:58
[10/19 14:24:38 d2.evaluation.evaluator]: Inference done 126/150. 6.7042 s / img. ETA=0:03:48
[10/19 14:24:48 d2.evaluation.evaluator]: Inference done 127/150. 6.7035 s / img. ETA=0:03:39
[10/19 14:25:06 d2.evaluation.evaluator]: Inference done 128/150. 6.7039 s / img. ETA=0:03:31
[10/19 14:25:14 d2.evaluation.evaluator]: Inference done 129/150. 6.7033 s / img. ETA=0:03:21
[10/19 14:25:29 d2.evaluation.evaluator]: Inference done 130/150. 6.7031 s / img. ETA=0:03:12
[10/19 14:25:47 d2.evaluation.evaluator]: Inference done 131/150. 6.7027 s / img. ETA=0:03:04
[10/19 14:26:06 d2.evaluation.evaluator]: Inference done 132/150. 6.7028 s / img. ETA=0:02:55
[10/19 14:26:24 d2.evaluation.evaluator]: Inference done 133/150. 6.7033 s / img. ETA=0:02:47
[10/19 14:26:35 d2.evaluation.evaluator]: Inference done 134/150. 6.7030 s / img. ETA=0:02:37
[10/19 14:26:49 d2.evaluation.evaluator]: Inference done 135/150. 6.7027 s / img. ETA=0:02:28
[10/19 14:27:00 d2.evaluation.evaluator]: Inference done 136/150. 6.7023 s / img. ETA=0:02:18
[10/19 14:27:15 d2.evaluation.evaluator]: Inference done 137/150. 6.7017 s / img. ETA=0:02:09
[10/19 14:27:24 d2.evaluation.evaluator]: Inference done 138/150. 6.7015 s / img. ETA=0:01:58
[10/19 14:27:38 d2.evaluation.evaluator]: Inference done 139/150. 6.7007 s / img. ETA=0:01:49
[10/19 14:27:48 d2.evaluation.evaluator]: Inference done 140/150. 6.7001 s / img. ETA=0:01:39
[10/19 14:27:59 d2.evaluation.evaluator]: Inference done 141/150. 6.6993 s / img. ETA=0:01:29
[10/19 14:28:07 d2.evaluation.evaluator]: Inference done 142/150. 6.6987 s / img. ETA=0:01:19
[10/19 14:28:15 d2.evaluation.evaluator]: Inference done 143/150. 6.6983 s / img. ETA=0:01:09
[10/19 14:28:24 d2.evaluation.evaluator]: Inference done 144/150. 6.6978 s / img. ETA=0:00:59
[10/19 14:28:32 d2.evaluation.evaluator]: Inference done 145/150. 6.6972 s / img. ETA=0:00:49
[10/19 14:28:43 d2.evaluation.evaluator]: Inference done 146/150. 6.6967 s / img. ETA=0:00:39
[10/19 14:28:54 d2.evaluation.evaluator]: Inference done 147/150. 6.6961 s / img. ETA=0:00:29
[10/19 14:29:02 d2.evaluation.evaluator]: Inference done 148/150. 6.6949 s / img. ETA=0:00:19
[10/19 14:29:13 d2.evaluation.evaluator]: Inference done 149/150. 6.6945 s / img. ETA=0:00:09
[10/19 14:29:21 d2.evaluation.evaluator]: Inference done 150/150. 6.6932 s / img. ETA=0:00:00
[10/19 14:29:21 d2.evaluation.evaluator]: Total inference time: 0:23:56.182427 (9.904706 s / img per device, on 2 devices)
[10/19 14:29:21 d2.evaluation.evaluator]: Total inference pure compute time: 0:16:10 (6.693243 s / img per device, on 2 devices)
[10/19 14:29:22 d2.evaluation.text_evaluation]: Saving results to ./output/inference/text_results.json
An invalid detection in temp_det_results/0000068.txt line 0 is removed ...
An invalid detection in temp_det_results/0000240.txt line 0 is removed ...
Calculated!
"E2E_RESULTS: precision: 0.0, recall: 0.0, hmean: 0"
"DETECTION_ONLY_RESULTS: precision: 0.6, recall: 0.0040650406504065045, hmean: 0.008075370121130552"
Calculated!
"E2E_RESULTS: precision: 0.0, recall: 0.0, hmean: 0"
"DETECTION_ONLY_RESULTS: precision: 0.5384615384615384, recall: 0.0031616982836495033, hmean: 0.006286484059272564"
[10/19 14:29:23 d2.engine.defaults]: Evaluation results for totaltext_test in csv format:
[10/19 14:29:23 d2.evaluation.testing]: copypaste: Task: DETECTION_ONLY_RESULTS
[10/19 14:29:23 d2.evaluation.testing]: copypaste: precision,recall,hmean
[10/19 14:29:23 d2.evaluation.testing]: copypaste: 0.6000,0.0041,0.0081
[10/19 14:29:23 d2.evaluation.testing]: copypaste: Task: None-E2E_RESULTS
[10/19 14:29:23 d2.evaluation.testing]: copypaste: precision,recall,hmean
[10/19 14:29:23 d2.evaluation.testing]: copypaste: 0.0000,0.0000,0.0000
[10/19 14:29:23 d2.evaluation.testing]: copypaste: Task: Strong-E2E_RESULTS
[10/19 14:29:23 d2.evaluation.testing]: copypaste: precision,recall,hmean
[10/19 14:29:23 d2.evaluation.testing]: copypaste: 0.0000,0.0000,0.0000

还有个问题，在预训练中文数据时，后面有个验证，显示效果非常差，这是怎么回事呀

权是海 @.***

------------------ 原始邮件 ------------------ 发件人: "权是海" @.>; 发送时间: 2023年9月22日(星期五) 中午11:07 @.@.>; @.>; 主题: 回复： [mxin262/SwinTextSpotter] batch_size (Issue #102)

非常感谢您的回复对于第2个问题，IMS_PER_BATCH和REC_HEAD下面的BATCH_SIZE存在比例关系嘛，因为我调整IMS_PER_BATCH为2这个时候，不知道后面识别的BATCH_SIZE应该设置为多少

权是海 @.***

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年9月22日(星期五) 中午11:03 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [mxin262/SwinTextSpotter] batch_size (Issue #102)

1）中文的类别更多，识别分辨率更大了，所以占用显存更多； 2）这是因为最大的文本数量限制了只能300个，调整识别头BATCH_SIZE之后，出现这个报错是因为图片中没有文本实例，可以继续训练的， 3）可以试试把文件先下载好，根据readme来评测

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

Oct 19 '23 07:10 Huangqqqhhh

中文数据只能上传到官网来评估

Oct 23 '23 06:10 mxin262

你好，在测试ic15时，发现，

category	#instances
text	0

权是海 @.***

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年10月23日(星期一) 下午2:51 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [mxin262/SwinTextSpotter] batch_size (Issue #102)

中文数据只能上传到官网来评估

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: @.***>

Oct 23 '23 07:10 Huangqqqhhh

SwinTextSpotter SwinTextSpotter copied to clipboard

batch_size

SwinTextSpotter
SwinTextSpotter copied to clipboard