why unsup_loss is 0.000

[yangzhaojason]

[MendelXu]

It is 0 at the beginning as no pseudo box is generated. However, it should be greater than 0 after several iterations.

[yangzhaojason]

box
About how many iterations will there be greater than 0

[yangzhaojason]

另外，我感觉是不是加载一个预训练模型会加速模型的收敛？感觉很多轮依然是0

[MendelXu]

box
About how many iterations will there be greater than 0

If you are using ms coco, it should increase quickly and be greater than 0 after about 1000~2000 iterations. For other datasets, I am not sure.

另外，我感觉是不是加载一个预训练模型会加速模型的收敛？感觉很多轮依然是0

Yes. adding load_from='{YOUR_MODEL_PATH}' in your config file should be ok.

[yangzhaojason]

谢谢，我尝试一下

[plalexx]

Noticed same problem while training on custom dataset even after large number of iterations. Have you figured what was the issue?

[MendelXu]

Maybe you can try to train a supervised model to see the supervised performance and have an impression of how the predictions look like.

[JLC7]

Same I get 0 unsup_loss bbox even though I used load_from with pretrained model. Sometimes it starts at a small value but eventually goes to zero and during evaluation there are no boxes. I had to set by_probto zero false according to #62 to fix the assert len(indices) == len(self).

[MendelXu]

I am not sure what the problem is. Could you try to train a supervised model and test it on your images to see whether the dataset and model are ready?

[JLC7]

Yes I trained a model using the a config in config/baseline/ and it seems to do well:

 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.938                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.843                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.610                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.683                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.797                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.737                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.658                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.732                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.835

When I try with soft teacher it is all zero (i.e. no boxes)

[MendelXu]

Is it possible for you to upload the log and the config file here?

[wjm202]

Yes I trained a model using the a config in config/baseline/ and it seems to do well:

 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.938                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.843                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.610                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.683                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.797                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.737                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.658                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.732                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.835

When I try with soft teacher it is all zero (i.e. no boxes)

do you solve this problem?I've been stuck with the same problem for days.

[shen0526]

Yes I trained a model using the a config in config/baseline/ and it seems to do well:

 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=1000 ] = 0.938                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=1000 ] = 0.843                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.610                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.683                                                                                                                                                                
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.797                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=300 ] = 0.737                                                                                                                                                                 
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=1000 ] = 0.737                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.658                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.732                                                                                                                                                                
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.835

When I try with soft teacher it is all zero (i.e. no boxes)

do you solve this problem？

[curiosity654]

I also encountered a similar problem, and I find that the bbox score of the teacher model will soon decrease to below the pseudo label threshold within a few iterations, which is strange.

[curiosity654]

I also encountered a similar problem, and I find that the bbox score of the teacher model will soon decrease to below the pseudo label threshold within a few iterations, which is strange.

Maybe it's because the threshold is set too high. The loss is zero because there's no pseudo label. For different datasets and models, the pseudo label thresholds are different.