youngjae-you comments

Results 73 comments of


                                            youngjae-you

Implement new conditional/optional Non Maximum Processor Pad Probe Handler

@rjhowell44 I'm looking at the code right now. I think it would be good to add the `class agostic option` using param to the `dsl_pph_nmp_new` I mentioned before. If `class...

Implement new conditional/optional Non Maximum Processor Pad Probe Handler

@rjhowell44 So, do you Implement that label_file is a parameter for parsing num_labels? The labels.txt used in osd is parsing what is in the infer config?

Implement new conditional/optional Non Maximum Processor Pad Probe Handler

Yes, I understand. Thank you.

Request to provide a service that can change the text offset at display render

dsl_ode_action_label_offset_new seems to have a bug

Is not l1 loss the smooth l1 mentioned in the paper?

Looks like he's taking the subgradient approach

Can this code be more gently predicated on the small box size?

@tensorturtle Hi! The results of the experiment showed that in my case, I lowered the variance for bbox prediction and it improved

When 20% pruning, one layer itself becomes zero, is this a problem?

@uyzhang 1. What is the criterion of being sufficient? 2. And although the map goes up steadily, l1 regularization can reverse the current sparsity calculation? Thank you for your reply

When 20% pruning, one layer itself becomes zero, is this a problem?

@uyzhang 1. I'm sorry, but where can I look that picture? Does it save every epoch? 2. During sparse training, my mAP rose stably from 0.5 to 0.65, for 20...

When 20% pruning, one layer itself becomes zero, is this a problem?

@uyzhang Thank you for your quick answerI 'll give it a try and tell you In addition, BN bias also does l1 regualization in this repo, and then 10 times......