Feng Li comments

Results 120 comments of


                                            Feng Li

Could guide me how to train this model in our own dataset that is in COCO format? step by step guide will be very helpful

If this dataset is already in COCO format, you need to change the number of classes according to your own dataset. I think that is enough to train your dataset.

MaskDINO performs poorly on single-class Object detection or Instance segmantation tasks?

We did not try training on single-class tasks. From my understanding, the performance of this setting should be good. Can you give more information?

MaskDINO support for CPU-only inference

Hey, we did not try CPU inference before, so we can not provide any suggestions now.

Provide colab or huggingface demo

Thanks for your suggestions. We will consider providing colabs in the future.

The environment configuration failed. Is there any way to reproduce it?

Hi, are you using the correct cuda and Pytorch version that matches your GPU arch?

Question about the segmentation branch

Yes. We use a 1/8 map in the encoder by default. The biggest map we use is 1/4 of the encoder (refer to our 5-scale model). It can improve the...

About the training on our datasets

The AP is pretty high, therefore the predictions are supposed to have very few duplicate predictions. As we are using the sigmoid focal loss, so for visualization, you need to...

About the training on our datasets

You need to write a visualization script, in which you manually select the queries with classification scores larger than 0.3, and then visualize them. 100 queries is good in your...

Change number of objects

Sorry for the late reply. Can you give more details about changing `NUM_OBJECT_QUERIES`?

Why is the config file name of Swin-L called R50?

Thanks. We will modify the name to swinL.