profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/angerhang/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.
angerhang Oxford, England hangyuan.xyz Self-supervised learning, ML4Health, wearables and epi

angerhang/biobankAccelerometerAnalysis 1

Extracting meaningful health information from large accelerometer datasets

activityMonitoring/capture24_neurips 0

Capture24: Activity recognition on a large activity tracker dataset collected in the wild

angerhang/angerhang.github.io 0

Personal site: angerhang.github.io

issue openedactivityMonitoring/capture24_neurips

Train more models

created time in a day

issue openedactivityMonitoring/capture24_neurips

Data cleaning experiment data-centric AI

Showing the mislabelled signals and why we can't just train on 200+ labels. Test this on three models: RF, pre-trained CNN and one more

created time in a day

issue openedactivityMonitoring/capture24_neurips

Varying the number of subjects' performance

Keep the test set fixed but vary the number of subjects in the training sets to see how much better than having 150 is compared to 10-15 which exiting datasets have.

created time in a day

issue openedactivityMonitoring/capture24_neurips

Inclusion of more datasets for baselines

2-3 more datasets and evaluate the datasets on all the models.

created time in a day

starteddaisukelab/cv_opt_flow

started time in 7 days

startedsniklaus/pytorch-liteflownet

started time in 7 days

startedprinceton-vl/RAFT

started time in 7 days

startedchuanenlin/optical-flow

started time in 7 days

issue commentactivityMonitoring/biobankAccelerometerAnalysis

Overclassification of sleep for low frequency aquisitions (5Hz)

Thanks for your questions. It is always nice to have external validation using our algorithm. 5Hz does seem a bit too low for sleep, making the device less sensitive to movement. To me 4/8g in terms of range doesn't make a big difference but something with 32Hz would likely help with your classification output.

@chanshing @R-Walmsley @aidendoherty do they also need to normalise the input for the random forest model by any chance?

SuperCorks

comment created time in 22 days

push eventactivityMonitoring/capture24_neurips

Shing Chan

commit sha 3e2bc972df33be4b80517f99ce6d00a4d29591ed

update conf to best found

view details

Shing Chan

commit sha 82d828f5a76c77af369ec9f12a3d123baf046b7b

resolve ckpt_path

view details

Shing Chan

commit sha 83db40a796d5ba8a125b05480a922098f272964a

fix trainval split

view details

Shing Chan

commit sha f97e66d7139a6defeb0350858e9dcc60cc11b6f6

update conf to best after train/val split fix

view details

Shing Chan

commit sha 7b4ac248a757dbfc12ae913533a80875c2c21756

add hmm

view details

angerhang

commit sha 96f0af90d23ae397e87236e78777aac29beb099e

Merge branch 'master' into cnn_lstm

view details

angerhang

commit sha 643d1bc75243598872d97907d096fffcf9881e72

include cnn+hmm

view details

push time in a month

push eventactivityMonitoring/capture24_neurips

angerhang

commit sha 8266566a5267bd2cdbad394069b8d65ee2e6869d

update doc

view details

push time in a month

push eventactivityMonitoring/capture24_neurips

angerhang

commit sha 6dd46a985cc1b396799f81d47c65abd6c3e39f0a

include ssl and cnn_lstm

view details

push time in a month

PR opened activityMonitoring/capture24_neurips

cnn_lstm working

Add cnn_lstm model and the relevant reporting.

To do: Produce the evaluation metrics on the test set when the model is done training.

+1122 -12

0 comment

15 changed files

pr created time in a month

push eventactivityMonitoring/capture24_neurips

angerhang

commit sha 5935aa33ee8bed4e2844963ed8ae9ce759412227

cnn_lstm working

view details

push time in a month

create barnchactivityMonitoring/capture24_neurips

branch : cnn_lstm

created branch time in a month

startedabhinavs/moonwalk

started time in a month

startedalbermax/innvestigate

started time in 2 months

pull request commentactivityMonitoring/biobankAccelerometerAnalysis

Fall back for downloadDataModel

Hi @angerhang - this still seems to fail for the same reason. I wonder if there are any useful nuggets of information from here? https://github.community/t/unexpected-step-failure-on-failing-bash-function-statement/144660

Actually NVM, I didn't include all the download properly. Should be ready to go now.

angerhang

comment created time in 2 months

push eventangerhang/biobankAccelerometerAnalysis

angerhang

commit sha a355fce2d494acb6d00504a5bb3a933eaf69c50b

Include all the models to download

view details

push time in 2 months

push eventangerhang/biobankAccelerometerAnalysis

angerhang

commit sha ee2cbf98146cf1776fb0a75d5525f099afce7986

Correct certificate check in model download

view details

push time in 2 months

startedjacobgil/pytorch-grad-cam

started time in 2 months

startedjphall663/awesome-machine-learning-interpretability

started time in 2 months

startedchr5tphr/zennit

started time in 2 months

PR opened activityMonitoring/biobankAccelerometerAnalysis

Fall back for downloadDataModel

This should fix #170 allowing --no-check-certificate to be used when the certificate cannot be verified.

All the CI tests should pass with this change.

+6 -6

0 comment

1 changed file

pr created time in 2 months

create barnchangerhang/biobankAccelerometerAnalysis

branch : fix-util

created branch time in 2 months

issue commentactivityMonitoring/biobankAccelerometerAnalysis

`cwa_regression` fails due to missing model resource

It is so decided that we will first try download the model without --no-check-certificate. If it fails, then we will try --no-check-certificate flag.

A patch is coming up soon.

angerhang

comment created time in 2 months

issue commentactivityMonitoring/biobankAccelerometerAnalysis

`cwa_regression` fails due to missing model resource

@aidendoherty who did you suggest that we should contact to double check the domain certificate? Is it the NDPH IT?

angerhang

comment created time in 2 months

issue commentactivityMonitoring/biobankAccelerometerAnalysis

`cwa_regression` fails due to missing model resource

After I re-run the tests, the issue persists. When I am running the download script on my local machine, I also have the same issue.

Resolving gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)... 163.1.206.75
Connecting to gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)|163.1.206.75|:443... connected.
ERROR: cannot verify gas.ndph.ox.ac.uk's certificate, issued by ‘CN=QuoVadis Global SSL ICA G3,O=QuoVadis Limited,C=BM’:
  Unable to locally verify the issuer's authority.
To connect to gas.ndph.ox.ac.uk insecurely, use `--no-check-certificate'.

It appears to be related to the domain certificate as reported earlier in https://github.com/activityMonitoring/biobankAccelerometerAnalysis/issues/153.

I also tried the download script on VM, the download time out on

--2021-07-15 06:23:11--  http://gas.ndph.ox.ac.uk/aidend/accModels/willetts-may20.tar
Resolving gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)... 163.1.206.75
Connecting to gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)|163.1.206.75|:80... 

failed: Connection timed out.
Retrying.

Likely it is not an issue specific to Mac but making all OS unstable for resource download.

Can @chanshing or @R-Walmsley reproduce the same issues?

angerhang

comment created time in 2 months

issue openedactivityMonitoring/biobankAccelerometerAnalysis

`cwa_regression` fails due to missing model resource

In our CI tests, cwa_regression will fail from time to time due to the following error:

--2021-07-14 15:37:34--  https://gas.ndph.ox.ac.uk/aidend/accModels/walmsley-jan21.tar
5179
Resolving gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)... 163.1.206.75
5180
Connecting to gas.ndph.ox.ac.uk (gas.ndph.ox.ac.uk)|163.1.206.75|:443... connected.
5181
Unable to establish SSL connection.
5182
URL transformed to HTTPS due to an HSTS policy

It sometimes fails using python 3.6 , 3.7, 3.8. Pretty annoying not sure what is causing this but someone will need to investigate this. I have started the CI tests for some of the current PRs.

created time in 2 months