doctorkey / c-3po

[PR] How to Reduce Change Detection to Semantic Segmentation

Hi there 👋. Welcome to Guo-Hua Wang's profile!

Guo-Hua Wang is currently a researcher and developer for deep learning. He focuses on:

• Multi-model Learning:
  • Knowledge Distillation: LSHFM.singleclassification, LSHFM.multiclassification, LSHFM.detection
  • Model Compression/Acceleration: Practise
• Few-label Learning:
  • Semi-Supervised Learning: R2D2.pytorch
  • Few-shot Learning: Practise
• Image Related Tasks:
  • Image Classification: LSHFM.singleclassification, LSHFM.multiclassification
  • Object Detection: LSHFM.detection
  • Semantic Segmentation: C-3PO
  • Change Detection: C-3PO
  • Image Retrieval: DIGIX_2020_B_image_retrieval, CBIR, image_retrieval_demo
  • Image Compression: DCVC

💼 Projects

📅 Dates	🎁 Projects	⭐ Stars	📚 Forks	🛎 Issues	📬 Pull requests
2022.11-2023.3	Practise
2023.1-2023.3	njuthesis
2022.3-2022.10	DCVC
2021.7-2021.11	C-3PO
2020.8-2020.11	DIGIX_2020_B_image_retrieval
2020.1-2020.8	LSHFM.singleclassification
2020.1-2020.8	LSHFM.multiclassification
2020.1-2020.8	LSHFM.detection
2019.1-2019.12	R2D2.pytorch

📊 Github stats

c-3po's Issues

Train on the ChangeSim

Hello, I appreciate your efforts.

I want to train the model on the ChangeSim.
However, I cannot find the below files.
Could you provide the files?

    ├── dark_test_list.txt
    ├── dust_test_list.txt
    ├── idx2color.txt
    ├── test_list.txt
    ├── test_split.txt
    ├── train_list.txt
    └── train_split.txt

Device parameter

in the main train.py script you can pass the device param. but there is a lot of stuff cuda Is used for and can't use the CPU to do anything? is this meant to do stuff like cuda:4 or so?

Model weights?

Hi, congratulations on the excellent work! Where can be accessed the weights of the models presented in the results table?

Test the model on new image pairs (img1, img2) and identify the differences.

How can I test model on new images data.
inference code on custom dataset........

Using pretrained weights but getting poor metrics?

Hello, I use the pre-trained weight "resnet18_id_4_deeplabv3_VL_CMU_CD.pth" to run the shell command "resnet18_mtf_id_msf4_deeplabv3_cmu", but the result is very bad, the following is my running log and results, can you help me find the reason?

PS C:\Users\TEST\Desktop\C-3PO-main> c:; cd 'c:\Users\TEST\Desktop\C-3PO-main'; & 'C:\Users\TEST\anaconda3\envs\pc\python.exe' 'c:\Users\TEST.vscode\extensions\ms-python.python-2023.4.1\pythonFiles\lib\python\debugpy\adapter/../..\debugpy\launcher' '52990' '--' 'C:\Users\TEST\Desktop\C-3PO-main/src/train.py' '--test-only' '--save-imgs' '--model' 'vgg16bn_mtf_msf_deeplabv3' '--mtf' 'id' '--msf' '4' '--train-dataset' 'VL_CMU_CD' '--test-dataset' 'VL_CMU_CD' '--input-size' '512' '--resume' 'C:\Users\TEST\Downloads\vgg16bn_id_4_deeplabv3_VL_CMU_CD.pth'
Not using distributed mode
Namespace(train_dataset='VL_CMU_CD', test_dataset='VL_CMU_CD', test_dataset2='', input_size=512, randomflip=0.5, randomrotate=False, randomcrop=False, data_cv=0, model='vgg16bn_mtf_msf_deeplabv3', mtf='id', msf=4, device='cuda', batch_size=4, epochs=100, workers=16, loss='bi', loss_weight=False, opt='adam', lr_scheduler='cosine', lr=0.0001, warmup=False, momentum=0.9, weight_decay=0, print_freq=50, resume='C:\Users\TEST\Downloads\vgg16bn_id_4_deeplabv3_VL_CMU_CD.pth', pretrained='', start_epoch=0, eval_every=1, test_only=True, save_imgs=True, save_local=False, world_size=1, dist_url='env://', output_dir='C:/Users/TEST/Desktop/output\vgg16bn_mtf_msf_deeplabv3_VL_CMU_CD_0/2023-03-13_19-29-23', distributed=False)
Train Aug:
DATA AUG: resize (512, 512)
DATA AUG: random flip 0.5
VL_CMU_CD train: 1008
Test Aug:
DATA AUG: resize (512, 512)
VL_CMU_CD test: 354
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torch\utils\data\dataloader.py:554: UserWarning: This DataLoader will create 16 worker processes in total. Our suggested max number of worker in current system is 8 (cpuset is not taken into account), which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
warnings.warn(_create_warning_msg(
MSF: [128, 256, 512, 512]
MTF: mode: id kernel_size: 3
MTF: mode: id kernel_size: 3
MTF: mode: id kernel_size: 3
MTF: mode: id kernel_size: 3
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\models_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed
in the future, please use 'weights' instead.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\models_utils.py:223: UserWarning: Arguments other than a weight enum or None for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing weights=VGG16_BN_Weights.IMAGENET1K_V1. You can also use weights=VGG16_BN_Weights.DEFAULT to get the most up-to-date weights.
warnings.warn(msg)
Loss: bi
LR Scheduler: cosine
load from: C:\Users\TEST\Downloads\vgg16bn_id_4_deeplabv3_VL_CMU_CD.pth
load ret:
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torch\utils\data\dataloader.py:554: UserWarning: This DataLoader will create 16 worker processes in total. Our suggested max number of worker in current system is 8 (cpuset is not taken into account), which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
warnings.warn(_create_warning_msg(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
C:\Users\TEST\anaconda3\envs\pc\lib\site-packages\torchvision\transforms\functional.py:442: UserWarning: Argument 'interpolation' of type int is deprecated since 0.13 and will be removed in 0.15. Please use InterpolationMode enum.
warnings.warn(
Test: [ 0/354] eta: 5:26:46 Prec: 0.999 (0.999) Rec: 0.034 (0.034) Acc: 0.092 (0.092) F1score: 0.0656 (0.0656) time: 55.3849 data: 53.1827 max mem: 717
Test: [100/354] eta: 0:03:54 Prec: 1.000 (0.994) Rec: 0.188 (0.113) Acc: 0.269 (0.250) F1score: 0.3163 (0.1955) time: 0.3864 data: 0.0014 max mem: 721
Test: [200/354] eta: 0:01:39 Prec: 0.995 (0.996) Rec: 0.115 (0.111) Acc: 0.157 (0.236) F1score: 0.2069 (0.1886) time: 0.3468 data: 0.0016 max mem: 721
Test: [300/354] eta: 0:00:30 Prec: 0.994 (0.981) Rec: 0.127 (0.104) Acc: 0.169 (0.214) F1score: 0.2245 (0.1784) time: 0.4156 data: 0.0014 max mem: 721
Test: Total time: 0:03:13
Test: Total: 354 Metric Prec: 0.9834 Recall: 0.0997 F1: 0.1728
tensor(0.1728)

ModuleNotFoundError: No module named 'correlation_cuda'

When trying to run the train script I get the ModuleNotFoundError: No module named 'correlation_cuda' error. I have tried this on my local machine and on a server from lambdalabs but both don't seem to work and keep getting stuck on the correlation_cuda issue

Question about PCD dataset 5-fold cross validation

When loading the PCD dataset, select the Set by data-cv option to load the image from the pre-separated train/test folder.
Please explain the code or method for separating images into folders for the same performance comparison.

Model inference for new set of images

Hi and thanks for the amazing work!
I was wondering if you could give me any advice on how to perform the inference on my own set of images.
Thank you,
Laura

Can the model output the types of changes (disappearance, appearance, exchange)

Hello, thank you for bringing new ideas. May I ask if it is possible to distinguish the appearance, disappearance, and exchange of change targets from the output results of the model?