bubbliiiing/faster-rcnn-pytorch

未检测到任何目标。

dingyan-dy opened this issue · 4 comments

dingyan-dy commented

您好，我前面按照流程生成了自己的数据集，训练过程中出现下列现象，显示所有评估参数为0而且未检测到任何目标，想问下有没有人知道这可能是什么原因呀？
`Number of devices: 1
initialize network with normal type
Configurations:

| keys | values|

| classes_path | model_data/voc_classes_custom.txt|
| model_path | |
| input_shape | [640, 640]|
| Init_Epoch | 0|
| Freeze_Epoch | 50|
| UnFreeze_Epoch | 300|
| Freeze_batch_size | 4|
| Unfreeze_batch_size | 16|
| Freeze_Train | False|
| Init_lr | 0.01|
| Min_lr | 0.0001|
| optimizer_type | sgd|
| momentum | 0.9|
| lr_decay_type | cos|
| save_period | 5|
| save_dir | logs|
| num_workers | 1|
| num_train | 743|
| num_val | 83|

[Warning] 使用sgd优化器时，建议将训练总步长设置到50000以上。
[Warning] 本次运行的总训练数据量为743，Unfreeze_batch_size为16，共训练300个Epoch，计算出总训练步长为13800。
[Warning] 由于总训练步长为13800，小于建议总步长50000，建议设置总世代为1087。
Start Train
Epoch 1/300: 100%|██████████| 46/46 [05:55<00:00, 7.72s/it, lr=0.001, roi_cls=1.61, roi_loc=0.0249, rpn_cls=0.672, rpn_loc=3.64, total_loss=5.95]
Finish Train
Start Validation
Epoch 1/300: 100%|██████████| 5/5 [00:17<00:00, 3.49s/it, val_loss=4.6]
Finish Validation
Epoch:1/300
Total Loss: 5.947 || Val Loss: 4.604
Save best model to best_epoch_weights.pth
Start Train
Epoch 2/300: 100%|██████████| 46/46 [04:39<00:00, 6.08s/it, lr=0.002, roi_cls=0.588, roi_loc=0.00784, rpn_cls=0.297, rpn_loc=1.15, total_loss=2.04]
Finish Train
Start Validation
Epoch 2/300: 100%|██████████| 5/5 [00:17<00:00, 3.43s/it, val_loss=1.44]
Finish Validation
Epoch:2/300
Total Loss: 2.037 || Val Loss: 1.437
Save best model to best_epoch_weights.pth
Start Train
Epoch 3/300: 100%|██████████| 46/46 [04:41<00:00, 6.13s/it, lr=0.005, roi_cls=0.372, roi_loc=0.0246, rpn_cls=0.106, rpn_loc=0.923, total_loss=1.43]
Epoch 3/300: 0%| | 0/5 [00:00<?, ?it/s<class 'dict'>]Finish Train
Start Validation
Epoch 3/300: 100%|██████████| 5/5 [00:17<00:00, 3.41s/it, val_loss=1.35]
Finish Validation
Epoch:3/300
Total Loss: 1.426 || Val Loss: 1.347
Save best model to best_epoch_weights.pth
Epoch 4/300: 0%| | 0/46 [00:00<?, ?it/s<class 'dict'>]Start Train
Epoch 4/300: 100%|██████████| 46/46 [04:33<00:00, 5.94s/it, lr=0.01, roi_cls=0.368, roi_loc=0.0269, rpn_cls=0.0975, rpn_loc=0.924, total_loss=1.42]
Finish Train
Start Validation
Epoch 4/300: 100%|██████████| 5/5 [00:17<00:00, 3.51s/it, val_loss=1.36]
Finish Validation
Epoch:4/300
Total Loss: 1.416 || Val Loss: 1.362
Start Train
Epoch 5/300: 100%|██████████| 46/46 [04:58<00:00, 6.49s/it, lr=0.01, roi_cls=0.363, roi_loc=0.0299, rpn_cls=0.095, rpn_loc=0.943, total_loss=1.43]
Finish Train
Start Validation
Epoch 5/300: 100%|██████████| 5/5 [00:18<00:00, 3.72s/it, val_loss=1.34]
Finish Validation
0%| | 0/83 [00:00<?, ?it/s]Get map.
100%|██████████| 83/83 [04:36<00:00, 3.33s/it]
Calculate Map.
loading annotations into memory...
Done (t=0.00s)
creating index...
index created!
Loading and preparing results...
DONE (t=0.00s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type bbox
DONE (t=0.15s).
Accumulating evaluation results...
DONE (t=0.06s).
Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.000
Get map done.
Epoch:5/300
Total Loss: 1.430 || Val Loss: 1.335
Save best model to best_epoch_weights.pth
Epoch 6/300: 0%| | 0/46 [00:00<?, ?it/s<class 'dict'>]Start Train
Epoch 6/300: 100%|██████████| 46/46 [04:59<00:00, 6.50s/it, lr=0.01, roi_cls=0.358, roi_loc=0.0236, rpn_cls=0.0947, rpn_loc=0.903, total_loss=1.38]
Epoch 6/300: 0%| | 0/5 [00:00<?, ?it/s<class 'dict'>]Finish Train
Start Validation
Epoch 6/300: 100%|██████████| 5/5 [00:17<00:00, 3.44s/it, val_loss=1.31]
Finish Validation
Epoch:6/300
Total Loss: 1.379 || Val Loss: 1.307
Save best model to best_epoch_weights.pth
Start Train
Epoch 7/300: 100%|██████████| 46/46 [04:23<00:00, 5.73s/it, lr=0.01, roi_cls=0.365, roi_loc=0.038, rpn_cls=0.095, rpn_loc=0.888, total_loss=1.39]
Epoch 7/300: 0%| | 0/5 [00:00<?, ?it/s<class 'dict'>]Finish Train
Start Validation
Epoch 7/300: 100%|██████████| 5/5 [00:13<00:00, 2.77s/it, val_loss=1.37]
Finish Validation
Epoch:7/300
Total Loss: 1.386 || Val Loss: 1.371
Start Train
Epoch 8/300: 100%|██████████| 46/46 [04:51<00:00, 6.34s/it, lr=0.01, roi_cls=0.363, roi_loc=0.0421, rpn_cls=0.0933, rpn_loc=0.929, total_loss=1.43]
Epoch 8/300: 0%| | 0/5 [00:00<?, ?it/s<class 'dict'>]Finish Train
Start Validation
Epoch 8/300: 100%|██████████| 5/5 [00:17<00:00, 3.47s/it, val_loss=1.35]
Finish Validation
Epoch:8/300
Total Loss: 1.428 || Val Loss: 1.354
Start Train
Epoch 9/300: 100%|██████████| 46/46 [04:49<00:00, 6.29s/it, lr=0.00999, roi_cls=0.359, roi_loc=0.0452, rpn_cls=0.094, rpn_loc=0.921, total_loss=1.42]
Epoch 9/300: 0%| | 0/5 [00:00<?, ?it/s<class 'dict'>]Finish Train
Start Validation
Epoch 9/300: 100%|██████████| 5/5 [00:17<00:00, 3.48s/it, val_loss=1.34]
Finish Validation
Epoch:9/300
Total Loss: 1.419 || Val Loss: 1.340
Start Train
Epoch 10/300: 100%|██████████| 46/46 [04:25<00:00, 5.78s/it, lr=0.00999, roi_cls=0.362, roi_loc=0.0391, rpn_cls=0.0923, rpn_loc=0.906, total_loss=1.4]
Finish Train
Start Validation
Epoch 10/300: 100%|██████████| 5/5 [00:17<00:00, 3.43s/it, val_loss=1.41]
Finish Validation
Get map.
100%|██████████| 83/83 [01:01<00:00, 1.35it/s]
Calculate Map.
未检测到任何目标。
Get map done.
Epoch:10/300
Total Loss: 1.399 || Val Loss: 1.411 `

Link20A commented

你好，请问你解决了吗

dingyan-dy commented
csycvtsbcsbbbycsb commented

你好，请问你解决了吗

你好，你现在解决了吗

maige1219 commented

后来因为名称不规范我重新从0000序号重命名后转化成voc格式

请问原来的命名格式是什么样子的？我似乎也遇到了类似的问题...loss值基本上为0，我怀疑是roboflow搞的鬼
epoch_loss
如图，roboflow会把下载的图片都加上后缀
image