vinairesearch / counting-detr Goto Github PK

I have a question about the annotation files. What's the difference between count_train and unseen_count_train?
From what I understand, count_train has same classes with count_val and count_test, while unseen_count_train and unseen_count_test don't share classes.
So, would it be appropriate to use the unseen_count_train.json for a few-shot setting?
Also, is the model in Tab.9 trained using unseen_count_train.json and evaluated using unseen_count_test.json?

can you share the pseudo gt of the train_data

could you share the pretrained weights in your paper to help reproduce the result.

How many epochs do you need to run to achieve better results? I ran about 200 epochs, but the predicted annotation is empty.

I only have instances_test.json and instances_val.json.

Hi, this is really an impressive work. I am interested in the work and wondering when the code and data would be released.

where can i get checkpoint？

Excuse me. I have tried the code about this work. However, what confused me is that the paper mentions to use CNN+FPN structure for feature extraction, but I didn't find the FPN in the code section. In fact, after I added FPN to the code myself, I found that the quality of the generated pseudo-labels decreased. Can you help me to explain this, thank you very much!

Cannot find the anchor_center.py under model file (in CountDETR_147_2nd_stage),could you please puublish it?

Hi VinAI research, great paper there for counting unknown objects!

Issue:
I have tried the following script to test the stage 1 with FSCD147 dataset.

1. This script is successfully run
2. this script raise "key error whs"

The error happens on stage 1, point 2 (as in the link above), where I get

KeyError: 'whs'

when trying to run this script on stage 1.

CMIIW, Could I know, is it a bug, unfinished label sharing, or something else?

Here I also attach the snapshot of the error, hope it helps.

Expected output: I could run the whole script in the weakly_supervise_fscd_147.sh without any error.

Thanks in advance~

there is no quaden/docker_images:pytorch_cuda102 image in docker hub

can you share the persudo lable file?thanks

the key value "id" are different in test_count.json and pseudo_lvis_test_cxcywh.json, and the dataloader can not work correctly.

python3.7: can't open file 'evaluate_val_and_test.py': [Errno 2] No such file or directory

quaden/docker_images:pytorch_cuda102 is not available