shanghua-gao / sod100k Goto Github PK

Can you report the speed of CSNet on a GTX 1080Ti GPU?

thanks for your good research. I faced a problem when I export to onnx as below:

thanks for your help
best regards

Hi, if I want to train csnet-L-x1 model, what should I do ?

Hi! First of all, thanks for sharing your great work on salient object detection!

I'd like to ask about skimage.transform.resize in ${REPO_ROOT}/CSNet/test.py.

In the code, it passes 'reflect' to mode for skimage.transform.resize.

                    img = resize(img, (cfg.TEST.IMAGE_H, cfg.TEST.IMAGE_W),
                                 mode='reflect',
                                 anti_aliasing=False)

And I found it means:

Pads with the reflection of the vector mirrored on the first and last values of the vector along each axis.

Is there any reason you choose 'reflect' rather than others like 'empty'?

Actually, I usually handle object detectors and the most repos I've seen just zero-pad to keep aspect ratios.

Have you experienced a performance gap between 'reflect' and others like 'empty'?

Can you provide the saliency maps of yours?

hello, i have trained my model but when runnng test.py am getting size mismatchsize mismatch error.

RuntimeError: Error(s) in loading state_dict for CSNet:
        size mismatch for stage0.0.conv1x1.conv.weight: copying a param with shape torch.Size([40, 3, 3, 3]) from checkpoint, the shape in current model is torch.Size([23, 3, 3, 3]).

when run "python test.py --config configs/csnet-L-x2.yml",I met this error. The 'yacs is a packge written by yourself or come with python. Please help me. Thank you and best wish.
Traceback (most recent call last):
File "test.py", line 9, in
from configs import cfg
File "/home/sln/SOD/SOD100K-master/CSNet_training/configs/init.py", line 2, in
from .defaults import _C as cfg
File "/home/sln/SOD/SOD100K-master/CSNet_training/configs/defaults.py", line 2, in
from yacs.config import CfgNode as CN
ModuleNotFoundError: No module named 'yacs

looking forward to the code

Thank you for your great work!
I tried to do some test, but find there was not right link for the pre-trained model.
Can you fix this problem?

Hi, thanks for your wonderful work!
I met a size not match error when I try a bigger batch size. However, when I used my own dataset, it works. Do I ignore something about the DUTs dataset?

FileNotFoundError: [Errno 2] No such file or directory: "CSNet/results/csnet-L-x2/FmeasureResults_csnet-L-x2_['DUTS-TE', 'ECSSD'].txt"

It will be open source ?

0557.jpg is a gray-scale image in ECSSD dataset. when runing CSNet‘s testing code

python test.py --config configs/csnet-L-x2.yml

error will occur

Traceback (most recent call last):
File "test.py", line 135, in
main()
File "test.py", line 53, in main
test(model, cfg.TEST.DATASETS, loadepoch)
File "test.py", line 91, in test
img = np.transpose((img - mean) / std, (2, 0, 1))
ValueError: operands could not be broadcast together with shapes (224,224) (3,)

May be you can judge the gray-scale image & convert it into RGB image

img = skimage.img_as_float(
    io.imread(os.path.join(img_dir, img_name)))
if len(img.shape) == 2:
    img = skimage.color.gray2rgb(img)

源码的block是bottle2neck，论文的是ilblock

Hello,

which labeling tool do you use to create the ground truth masks ?

Hi! First of all, thanks for sharing your great work on salient object detection!

When i reading the source code 【SOD100K/CSNet_training/model/csnet.py/build_model]】
I cannot find the implementation of the function 'redefine_model' in line918.

please give me some help!

导致net.load_state_dict()报错

RuntimeError: Error(s) in loading state_dict for CSFNet:
Unexpected key(s) in state_dict: "base.fc.weight", "base.fc.bias".

请问有没有 ResNet50 的版本以及相应得到的saliency maps，而不是 Res2Net50

how can i export CSNet to onnx?

You only provide the saliency maps of CSF+Res2Net. Can you provide the saliency maps of CSNet?

你把一个256 * 256 * 3 * 3的卷积分成5个51 * 256 * 3 * 3的卷积、再concat。意义何在？
理论上没区别啊