lkevinzc / dance Goto Github PK
View Code? Open in Web Editor NEWCodes for "DANCE: A Deep Attentive Contour Model for Efficient Instance Segmentation", WACV2021
Codes for "DANCE: A Deep Attentive Contour Model for Efficient Instance Segmentation", WACV2021
Hi, I preprocessed COCO2017 dataset with python datasets/register_coco_edge.py
. But when I trained this network with python train_net.py --num-gpus 1 --config-file configs/Dance_R_50_3x.yaml
, I still faced a problem which said:
'ERROR [05/08 10:49:58 d2.engine.train_loop]: Exception during training:
Traceback (most recent call last):
File "/home/caoyang/detectron2/detectron2/engine/train_loop.py", line 132, in train
self.run_step()
File "/home/caoyang/detectron2/detectron2/engine/train_loop.py", line 214, in run_step
loss_dict = self.model(data)
File "/home/caoyang/anaconda3/envs/dance/lib/python3.6/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/caoyang/dance/core/modeling/edge_snake/dance.py", line 140, in forward
features, proposals, (gt_sem_seg, [gt_instances, images.image_sizes])
File "/home/caoyang/anaconda3/envs/dance/lib/python3.6/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/caoyang/dance/core/modeling/edge_snake/edge_det.py", line 270, in forward
_, poly_loss = self.refine_head(snake_input, None, targets[1])
File "/home/caoyang/anaconda3/envs/dance/lib/python3.6/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/home/caoyang/dance/core/modeling/edge_snake/snake_head.py", line 1881, in forward
training_targets = self.compute_targets_for_polys(gt_instances, image_sizes)
File "/home/caoyang/dance/core/modeling/edge_snake/snake_head.py", line 1232, in compute_targets_for_polys
init_ex_targets = torch.stack(init_ex_targets, dim=0)
RuntimeError: stack expects a non-empty TensorList'
I guess the reason is that there is no target in the picture, or the target is not marked. And would you like to tell me how to solve this problem.
Could you please offer an source mmdet version of DANCE?
Hi,
I really like your work!
I would like to apply dance to another dataset , so I need to train from scratch , it this possible ?
Also, do I need to apply any filters on the images or normalization techniques before training ?
Alternatively, I would like to use your algorithm for refining another segmentation model, hence using the masks generated from the 1st algorithm as seeds for DANCE ,so it cant make refinements by finding the "real contours".
Could option 1 or 2 (or both) is possible ? and how to do so ?
Thanks a lot,
Hi,
Thank you for your great work on Dance! I am trying to better understand this work so I want to train this model. I noticed that you have answered a issue about this, and I want to know if the config in https://github.com/lkevinzc/dance/blob/master/core/config/defaults.py is the same as the config you used in pre-trained model you provided? I noticed you set _C.MODEL.SNAKE_HEAD.ATTENTION = False
and _C.MODEL.SNAKE_HEAD.NEW_MATCHING = False
.
Sorry to bother you if I have misunderstood the model.
Hi, so sorry to bother you. Is there any code for the edge map generation from data of SBD? I only find coco and cityscapes in the datasets folder. Thanks.
I have read your paper, and it mentioned you re-implemented DeepSnake with FCOSas the object detector for fair comparison. So
can you tell me which part re-implemented DeepSnake with FCOS in your code? Thanks !
Hi, thanks for your excellent codes. Now, I want to train the basic dsnake_head. Would you like to provide the config(.yaml) file of this model? Look forward to your reply!
Traceback (most recent call last):
File "train_net.py", line 14, in
import core.data # noqa
File "/public/home/archer/dance/core/init.py", line 1, in
from core import modeling
File "/public/home/archer/dance/core/modeling/init.py", line 1, in
from .fcos import FCOS
File "/public/home/archer/dance/core/modeling/fcos/init.py", line 1, in
from .fcos import FCOS
File "/public/home/archer/dance/core/modeling/fcos/fcos.py", line 10, in
from core.layers import DFConv2d, IOULoss
File "/public/home/archer/dance/core/layers/init.py", line 4, in
from .extreme_utils import _ext as extreme_utils
ImportError: cannot import name '_ext' from 'core.layers.extreme_utils' (/public/home/archer/dance/core/layers/extreme_utils/init.py)
And I find there's nothing in init.py
Hi,
Sorry for bothering you, and thank you for the excellent work on DANCE! I am trying to better understand DANCE's improvements over DeepSnake, and I would like to reproduce more of DANCE's results. Do you have the COCO training script and pre-trained model for DANCE with CenterNet as the detector?
Thank you so much.
Hi, so sorry to bother you, but where is the training usage? And I noticed that some necessary file were missing, such as: core/modeling/backbone/mobilenet.py & vovnet.py, the panopticapi package, where can I find them?
when i training my own dataset ,have the problem about:
AssertionError: Attribute 'thing_classes' in the metadata of 'coco_2017_train_edge' cannot be set to a different value!
['person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light', 'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow', 'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee', 'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard', 'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple', 'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch', 'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone', 'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear', 'hair drier', 'toothbrush'] != ['building']
please give a suggestion~
Thanks for your great work. I have a question that why do you reproduce your code on detectron2 when training on coco, instead of directly using 'snake' code? Since files of coco dataset (dataset and evaluation for coco) have already existed in 'snake' (though snake does not use them).
Thanks for your work and code! An error occurs when I am running the baseline snake model using the command:
CUDA_VISIBLE_DEVICES=0,1 python train_net.py --num-gpus 2 --config-file configs/Dsnake_R_50_1x.yaml
File "/data/yinyf/dance/core/modeling/dsnake_baseline/dsnake_head.py", line 190, in forward
_, losses = self.refine_head(features["p2"], None, targets)
File "/home/fengh/miniconda3/envs/dance/lib/python3.6/site-packages/torch/nn/modules/module.py", line 532, in call
result = self.forward(*input, **kwargs)
File "/data/yinyf/dance/core/modeling/dsnake_baseline/dsnake_head.py", line 771, in forward
training_targets = self.compute_targets_for_polys(targets)
File "/data/yinyf/dance/core/modeling/dsnake_baseline/dsnake_head.py", line 638, in compute_targets_for_polys
init_sample_locations = torch.stack(init_sample_locations, dim=0)
RuntimeError: stack expects a non-empty TensorList
How can I deal with it?
Hi
When I run the following command,
python train_net.py --config-file configs/Dance_R_101_3x.yaml --eval-only MODEL.WEIGHTS ./output/r101_3x_model_final.pth
DANCE model predicts box coordinates and classes but does not produce a segmentation mask, as shown below.
How do I get the segmentation?
Also, which code did you use to visualize the predictions, like the figures in your DANCE paper?
Thank you for your help in advance.
First of all, thank you for your excellent work.I have successfully tested your model so far, but when I tried to train the model so that I could observe the flow of data, I ran into a problem: I only have one piece of 1080Ti(11G), and when I ran it, the video memory was exceeded. What parameters should I adjust if I want to train on a video card?Could you give me some advice?Thank you very much!
Hi, I have noticed that DeepSnake adopt the two-stage training strategies on cityscapes dataset , which train the detector alone firstly and then train the detector and snake branches simultaneously. And I also noticed that Dance are trained using SGD. So I want to know if Dance has adopted a two-stage training strategy, or directly trained end-to-end.
Thanks for your reply!
Hello! Thanks for you great work. Can you give some help about using COCO dataset in the snake branch? Because I want to use my own dataset which is the form of the COCO dataset to train the DANCE network in the snake branch. Thankyou!
您好,我主要想请教一下在snake分支下该如何支持DANCE模型训练COCO数据集形式的数据呢?主要是在lib/datsets/coco路径下和lib/evaluators/coco路径下相关的dance.py程序如何编写呢?
Thank you for your great work,I have a question:
In the paper,you add a "tanh" activation function to the snake prediction before multiplying it by the object scale.Why do you choose the activation function?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.