profile
viewpoint
Tianfei Zhou tfzhou Computer Vision Laboratory, ETH Zurich Switzerland http://www.tfzhou.com/

tfzhou/ContrastiveSeg 323

ICCV2021 (Oral) - Exploring Cross-Image Pixel Contrast for Semantic Segmentation

tfzhou/MATNet 136

Motion-Attentive Transition for Zero-Shot Video Object Segmentation (AAAI2020)

hlesmqh/WS3D 87

Official version of 'Weakly Supervised 3D object detection from Lidar Point Cloud'(ECCV2020)

tfzhou/C-HOI 72

Cascaded Human-Object Interaction Recognition (CVPR2020)

tfzhou/MG-HumanParsing 55

Differentiable Multi-Granularity Human Representation Learning for Instance-Aware Human Semantic Parsing

Lixy1997/Group-WSSS 54

Group-Wise Semantic Mining for Weakly Supervised Semantic Segmentation

tfzhou/BINGObjectness 24

BING Objectness proposal estimator Matlab wrapper. More in http://mmcheng.net/bing/

tfzhou/FFIW 20

Face Forensics in the Wild

tfzhou/Detectron-with-tensorboard 6

FAIR's research platform for object detection research, implementing popular algorithms like Mask R-CNN and RetinaNet.

issue closedtfzhou/MATNet

光流图是否在DAVIS数据集上训练过?

你好,我想请问一下,准备数据环节,准备的光流图是直接用pytorch-pwc在其他光流数据集上训练得到的权重预测生成的吗,您有在DAVIS数据集上训练吗,如果有,可不可以提供一下预训练模型呢?

closed time in 3 days

huanglf714

startedlocuslab/qpth

started time in 6 days

fork tfzhou/mmsegmentation

OpenMMLab Semantic Segmentation Toolbox and Benchmark.

https://mmsegmentation.readthedocs.io/en/latest/

fork in 7 days

issue closedtfzhou/ContrastiveSeg

tSNE visualisation

Hi,

Thanks for your great work! I am wondering how you select feature embeddings for tSNE visualisation? Because for the dense pixel-level segmentation task, if we use feature embeddings of all the pixels, it will be too much.

Many thanks!

closed time in 7 days

YuemingJin

issue commenttfzhou/ContrastiveSeg

tSNE visualisation

Yes, using all pixels is not necessary. Instead, you can sample a small number of pixels for tsne visualization.

YuemingJin

comment created time in 7 days

issue closedtfzhou/ContrastiveSeg

Explanation of n_view

Hi, Thanks for your great work.

Could you help explain what is the n_view stands for? https://github.com/tfzhou/ContrastiveSeg/blob/310120712ad4b6ecea45c3ac1143118f959c86e4/lib/loss/loss_contrast.py#L47-L51

I understand the meaning of the total_classes and feat_dim, but it's difficult for me to understand the meaning of n_view here. Thanks.

closed time in 7 days

HenryPengZou

issue commenttfzhou/ContrastiveSeg

when training code will be available?

Hi, please refer to the script provided in the table of ReadMe for model training.

seyeeet

comment created time in 7 days

issue commenttfzhou/ContrastiveSeg

would you provide a command to run this code?Thanks a lot

Hi, please refer to the script provided in the table of ReadMe for model training.

Brandy0k

comment created time in 7 days

push eventLixy1997/Group-WSSS

Tianfei Zhou

commit sha 62653ef34ec12bdf6be05f926eb40b13b7bd8a4a

Update README.md

view details

push time in 7 days

push eventLixy1997/Group-WSSS

Tianfei Zhou

commit sha 95c8fb595c102d9f57372e056cf53985f9d8e6dd

Update README.md

view details

push time in 7 days

push eventLixy1997/Group-WSSS

Tianfei Zhou

commit sha 3eb7307720532a53b559476da0cf7a9d2020e5cb

Update README.md

view details

push time in 7 days

issue commenttfzhou/MATNet

光流图是否在DAVIS数据集上训练过?

No, the optical flow model is not fine-tuned on DAVIS. We directly use the off-the-shelf model.

huanglf714

comment created time in 12 days

issue commenttfzhou/ContrastiveSeg

论文相似

@nlp-tokio 感谢你的提醒。我们的文章是2021年1月28号上传到arXiv上的,并同时开源了代码,在2月23号也被机器之心进行了宣传。你指出的这篇文章与我们的文章极度相似,不论是idea,部分论文描述,甚至Figure都有模仿我们文章的痕迹。对此,我们也十分的震惊。我们会尽力弄清楚这件事情,再次感谢你的提醒。

nlp-tokio

comment created time in a month

issue commenttfzhou/ContrastiveSeg

Explanation of n_view

Hi, I admit that n_view is not a good name. Here it refers to the maximum number of pixels sampled for each class.

HenryPengZou

comment created time in a month

startedJunMa11/SOTA-MedSeg

started time in a month

issue commenttfzhou/ContrastiveSeg

Function _dequeue_and_enqueue() in train_contrastive.py

Hi jiyang, the function is designed for memory updating. The pixel and segment queues store different feature embeddings, pixel or segment (average embedding of all pixels in a region). These embeddings are taken as positive or negative samples of pixels in each mini-batch for contrastive loss computation.

Jiyang-Zheng

comment created time in a month

issue commenttfzhou/ContrastiveSeg

Why ResNet-101 + DeepLab-V3 baseline performs so bad in your repo?

Hi Mengzhang, I am sorry for that I did not have enough time to study this baseline model. I previously use openseg with pytorch0.4.1, and the baseline works well. But after upgrading to pytorch1.7, it shows bad performance as reported in my ReadMe. I guess it is due to the syncbn problem, and I may study this later.

MengzhangLI

comment created time in a month

issue commenttfzhou/ContrastiveSeg

Problem in the function "_dequeue_and_enqueue"

Very grateful for reporting this issue. This is indeed a bug, which is mainly due to my re-implementation of this part. I will update the code accordingly. Thanks.

eezywu

comment created time in a month

startedDuncanZauss/Keypoint_Communities

started time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha f7d7275691233d5338883b7d3f362b83701fe7fc

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha 0f24e9dffcd59b39b5b5a18407d8480a05fce0b8

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha 9262ae69303e26d5475b1fb8b8392a1151db7cb3

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha a98692e367c3b797423785e5853c3f9e5719bed1

Add files via upload

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha b78161c677c2b495f69452498abb7f5159e512fa

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha a4afeebb857c509f86677b67866d2a37cfae17e7

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha 7defbf0079d40199daa502f0914c2a6304a293b2

Update README.md

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha 62e5f16906f1a704aca44d2b5088892c7772b077

Add files via upload

view details

push time in 2 months

push eventtfzhou/FFIW

Tianfei Zhou

commit sha 030a56a61bb0052f346e9e59376a727e4f6932a2

Update README.md

view details

push time in 2 months

push eventtfzhou/ContrastiveSeg

Tianfei Zhou

commit sha 7be326ae9127a847558584eac489d4858adf7cb1

Update README.md

view details

push time in 2 months

more