sentient07 / headhunter--t Goto Github PK
View Code? Open in Web Editor NEWThis repository is the official implementation of HeadHunter-T, the head tracker discussed in the CVPR paper, mentioned herewith.
This repository is the official implementation of HeadHunter-T, the head tracker discussed in the CVPR paper, mentioned herewith.
Could you please explain, how to train this dataset? or how to download the trained models "FT_R50_epoch_24.pth"?
Unlike in the HeadHunter
repo, there is no license file in this repository. Could you please add it or specify the license in any other way?
I put regression_thresh to 0 in the config_reproduce.yaml file, otherwise, I was ending up with this error if I run run_mot.py.
RuntimeError: boxes should be a 2d tensor, got 1D
Do you have an explanation for this ?
I try to test the run_new.py with my own images, but the error occurs that ModuleNotFoundError: No module named 'config.retina_cfg'. I find that config fold is not file named retian_cfg.
Thanks for your wonderful work!
I'd like to test my tracking method on detection result of HT-21 test data . And I'd like to use HeadHunter predict the detections.
So I'm wondering what data was used to train your pretrained model. Did you use the HT-21 test data?
Thanks!
Step 13/27 : ADD retina_env.yml /tmp/environment.yml
ADD failed: file not found in build context or excluded by .dockerignore: stat retina_env.yml: file does not exist
Hello, can you provide the retina_env.yml file?
Have you published the pre-training model and we would like to test it on our data set to see how it works
Where can I download the pretrained model weight of headhunter?
when I try to run the tracker on a CroHD dataset(training set) and **set the "use_public" option in the 'det_cfg' to 'False',**the ValueError said that "Anchors should be Tuple[Tuple[int]] because each feature map could potentially have different sizes and aspect ratios. "
I noticed that in the obj_detect.py,if det_cfg['median_anchor']:the program can choose different benchmark to import different anchors,so which benchmark I should set in the det_cfg when I want to run the tracker on a CroHD dataset?And I noticed that the benchmark should be associated with the anchor.py,too.
Another question is,if i set the "use_public" option in the 'det_cfg' to 'True',does it mean that I will use the data in det.txt which the dataset offers and the detector will not be used to detect the pedestrians or output anything.
Hello, on the Setup Instructions
section of the README.md file, it is stated that "Cuda 10.0 is needed if Docker is unavailable."
However, when I looked at line 45 of the env_super.yml
file, I find that a torch version with cuda 11.1 is installed (torch==1.8.1+cu111
).
As such, is cuda 10.0 or cuda 11.1 the requirement to run the model? Or is there something else that I'm not noticing?
Best regards,
Erick Platero
Hi, I am follow with interest your CroHD. It is very impressive. However, I have a question about the Table 4 in the paper "Tracking Pedestrian Heads in Dense Crowd". It shows the Headhunter--T method achieves a MOTA 63.6 and IDF1 57.1, while on MOT challenge https://motchallenge.net/results/Head_Tracking_21/ the Headhunter--T method only achieves a MOTA 57.8 and IDF1 53.9. Could you please tell me why there is an inconsistency between the two results and where is the gap. Thank you!
For the 4 step:
HeadHunter - the head detector to be installed as a python package and the path to weights of pre-trained head detector.
I can not install it.
this link is not available now, could you mind sharing the latest link to help me get FT_R50_epoch_24.pth
?
Hello there,
When i tried to run on a custoam dataset,am seeing the tracker ids are maintained in alternate frames for many people. For example if person1 has tracker ID 1 in frame1,ID 2 in frame2 then the ids are maintained for frame3,frame4,frame5,frame6 as ID1,ID2,ID1,ID2,ID1,ID2. This is weird,is it because of the nms ? Also all my head detections score are 1 when i am supplying to tracker.
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.