im willing to use this model by onnx.
can it be?

I want to know what operations make the edge maps finer.If you could spare some time to reply to me, I would greatly appreciate it.

What are the specific downstream tasks of edge images generated by edge detection? I noticed that you used it for sketch detection, are there any other specific applications?
Looking forward to your response.

Hi, I am working on another computer vsion field, I am interesting in your research, but When I use TEED as a part of my model, I found a error, It is normal during the train period, but something went wrong during my test period, When test image have different height and weight,here is the error:

I want to figure it out what happens, thanks!

hi, this code is very interesting, but i can not run this demo.
when i run python main.py --choose_test_data=-1, i have this wrong

Hello and thank you for your amazing works. I'm new to edge detection field so I would like to ask you a question. When I inspect your code I find that the DoubleFusion layer, which output the final edge map of the network, ended with a Fmish activation layer. When I plot the function, it show that most part have near-zero or positive value (except a very a narrow range near zero which can output small negative values). And in the loss layer again tou put the output through a sigmoid so, basically what should be non-edge are now have values near 0.5, which doesn't make sense to me. Can you explain further about this choice?

My result is inconsistent with that of python reasoning.

I have set the following lines in main.py:
25: is_testing =False

training with BIPED

223: TRAIN_DATA = DATASET_NAMES[0]

I set the following settings in the dataset. py file:
'BIPED': {
'img_height': 720, #720 # 1088
'img_width': 1280, # 1280 5 1920
'test_list': 'test_pair.lst',
'train_list': 'train_rgb.lst', # Base augmentation
# 'train_list': 'train_pairB3.lst', # another augmentation
# 'train_list': 'train_pairB5.lst', # Last augmentation
'data_dir': '/share/home/datasets/BIPED/edges', # mean_rgb
'yita': 0.5,
'mean':BIPED_mean
#
},
When I run:python main.py,it appears the following problems:
Training details> ['TEED BIPED+BRIND-trainingdataLoader BRIND light AF -USNet--noBN xav init normal bdcnLoss2+cats2loss +DoubleFusion-3AF, AF sum RL= 0.0008 WD= 0.0002 image size = 300 adjust LR=[4] LRs= [8e-05] Loss Function= BDCNloss2 + CAST-loss2.py Tue Dec 19 15:28:32 2023 trained on BIPED']
Number of GPU's available: 4
Pytorch version: 1.13.1
Trainimage mean: [103.939, 116.779, 123.68, 137.86]
Test image mean: [104.007, 116.669, 122.679, 137.86]
Traceback (most recent call last):
File "main.py", line 524, in
main(args, train_info)
File "main.py", line 408, in main
arg=args
File "/share/home/huangsheng/current/TEED/dataset.py", line 478, in init
self.data_index = self._build_index()
File "/share/home/huangsheng/current/TEED/dataset.py", line 504, in _build_index
files = json.load(f)
File "/share/home/huangsheng/anaconda3/envs/pytorch/lib/python3.7/json/init.py", line 296, in load
parse_constant=parse_constant, object_pairs_hook=object_pairs_hook, **kw)
File "/share/home/huangsheng/anaconda3/envs/pytorch/lib/python3.7/json/init.py", line 348, in loads
return _default_decoder.decode(s)
File "/share/home/huangsheng/anaconda3/envs/pytorch/lib/python3.7/json/decoder.py", line 337, in decode
obj, end = self.raw_decode(s, idx=_w(s, 0).end())
File "/share/home/huangsheng/anaconda3/envs/pytorch/lib/python3.7/json/decoder.py", line 355, in raw_decode
raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
If you could tell me what the problem is, I would greatly appreciate it.

I would like to ask how to enhance the BIPED dataset, the download is only 258 sheets. It doesn't correspond to train_pair0.lst, can you share the code for dataset enhancement or the dataset after enhancement.

Hello, sir.
When I use your TEED model for training, I get the.pth file. However, the quality of the prediction results obtained by using the weight file I trained myself is very poor, while using the weight file in your source code can obtain good prediction results. I don 't know if there is a problem with the parameter setting. Moreover, I used the BIPED dataset to train 8 epochs in less than 5 minutes, while the training time mentioned in your article is 30 minutes. I suspect that something must have gone wrong. Can you tell me your training parameter settings ?
The first image is the prediction result obtained by using the weight file obtained by my own training, and the second is the prediction result obtained by using the weight file in your source code.
Hope to receive your reply, thank you very much !

Traceback (most recent call last):
File "f:\subject\code\TEED-main\main.py", line 18, in
from loss2 import *
File "f:\subject\code\TEED-main\loss2.py", line 3, in
from utils.AF.Fsmish import smish as Fsmish
ModuleNotFoundError: No module named 'utils.AF'

May I ask if you have a pre training model that has been well trained and has a good effect

I used my edge diagram to carry out the sketch recognition project mentioned in the paper, and in order to convert the obtained edge diagram into a suitable MAT file, I wrote a jpg to MAT file script, but the prediction effect was very poor, how did you solve it in the first place?Looking forward to your reply！

How to evaluate ODS/OIS and AP IQA in edge detection?

Thank you for your great works about edge detection named TEED.
using TEED, i can find edges conveniently. after find edges, i want to extract xy coordinates of edges.
is there any way to extract xy coordinates in your code? Or does opencv can do that?
any suggestions appreciate.
thank you.

Hey, I am planning on fine-tuning this model with a variety of subjects such as faces, animals, and logos. I decided on this model because it closely alligns with the desired output.

My dataset consists of 100 images, each with a resolution of 2500x2500 pixels. Given the specific nature of my project, these images are significantly larger and potentially more complex than those in the BIPED dataset originally used to train the TEED model.

Challenges & Questions
Data Augmentation & Preparation: I've followed the BIPED dataset's augmentation process for my dataset for the most part. Considering the size and diversity of my images, are there any specific augmentation techniques or preprocessing steps you would recommend?

Model Fine-tuning: Given the unique size (2500x2500) of my dataset's images and their content diversity, I will need to make adjustments to the training procedure outlined for the BIPED dataset. Could you provide guidance or recommendations on optimizing the training process for a dataset with such characteristics? Specifically, I'm interested in any modifications to the training parameters, input size handling, or batch processing to accommodate the high-resolution images.

Loss Functions: The current loss functions used for TEED should seems like it will work for my task without modifications. However, I'm open to suggestions if you think adjustments or additional loss functions could enhance the model's performance to my dataset.

Thank you for your time, hope to hear back!

I'm new to learning deep learning, so please forgive me if I'm asking a really stupid question.
I found that the BIPED I downloaded was very simple and did not have folders such as “aug” or “p1” as mentioned in the train_pair0.lst. My attempts to train the TEED model also failed. I was hoping someone could tell me the solution.
I would be very grateful if you could provide me with the correct documentation.

When reading a BIPED dataset in JSON format, an error is reported, as shown below

我有几个关于评价指标的问题想问一下：
1.请问作者是否用BIPDE的训练集测试过指标，如果测试过，可否说一下，想验证一下我的测试结果。
2.关于NMS的参数问题，不知道这些参数为何这么设置，不知作者BIPDE和UDED进行评价的时候用的参数是什么？（(tmp_edge,O,1,5,1.03,8)？）还有就是参数的不同对结果的影响大不大？
3.关于evaluation过程，有没有什么需要特别注意的？

您好，我下载了你在issue里面提供的evaluation代码，在gt和edges_pred下面均放了.png图片，但是无法跑通

Base_dir: results\biped_ted
Make sure you have data in GT/ edge-maps dirs, then Run