thurssq / tianchi04 Goto Github PK

I feel a little confused about the output of yolov4, which is a list of 3 arrays, looking like the boxes, object confidence and classification. However, when I look into the codes of postprocessing,

only the objectless is used as the confidence and doesn't match the comments.

Could you please explain a little bit or refer me to the original repo of yolov4?

Hello, I encountered the following issues while trying to reproduce your code using mmcv and mmdet. Could you please explain the reasons for this and suggest how to resolve them?

Traceback (most recent call last):
File "attack_1.py", line 146, in
mask = get_mask(data['img'], data['img_metas'], pixels)
File "attack_1.py", line 71, in get_mask
bbox, label = model2(return_loss=False, rescale=True, img=image, img_metas=meta)
ValueError: not enough values to unpack (expected 2, got 1)

Hi, first of all thank you for your hard work on this subject!
I am trying to generate adversarial examples using your patch generation method, but I am slightly confused with a few things.
First, is it required for me to train my own Yolov4 and FasterRCNN models, or would pretrained models be ok? (I am only using images with cars)
Also, is there any guide to go about with the patch generation?
Thank you once again!

hello,There is a problem when put some pictures in images. what is the format of the data set？

Can somebody give the mmcv --version that needs to be installed?

Is there one available? Inlcuding torch / torchvision / etc?

Thanks a lot!

could you add the file show that how to install and use it on my custom dataset? 3q

File "E:/workspace/Tianchi04/attack_2.py", line 72, in get_mask
bbox, label = model2(return_loss=False, rescale=True, img=image, img_metas=meta)
ValueError: not enough values to unpack (expected 2, got 1)

@THUrssq

I want to download weights not program from baidu. Please provide alternative link to some google drive, or any other storage

Hi,

I want to ask a question about the size of input image when inputing to faster rcnn. Why should we need to resize the size to 800x800? I did not see the requirements in the relevant literature or mmdetection. Could you give me some advice? Thank you in advance.