aiprogrammer / advmix Goto Github PK
View Code? Open in Web Editor NEWOfficial code for our CVPR 2021 paper: "When Human Pose Estimation Meets Robustness: Adversarial Algorithms and Benchmarks".
License: MIT License
Official code for our CVPR 2021 paper: "When Human Pose Estimation Meets Robustness: Adversarial Algorithms and Benchmarks".
License: MIT License
Hi,
I wonder if can you make your pretrained model under AdvMix regime public?
I downloaded the raw images of COCO 2017 validation and MPII. The remaining time of executing make_dataset.sh
is very long, like 1000+ hrs. Is it correct?
Thanks for your nice work. I am running normal training. I think it is a simple baseline for normal training. So I think the following code (Line 44) should be changed from
for i, (input, target, target_weight, meta) in tqdm(enumerate(train_loader)):
data_time.update(time.time() - end)
outputs = model(input)
target = target[0].cuda(non_blocking=True)
target_hm = target
target_weight = target_weight.cuda(non_blocking=True)
loss = criterion(outputs, target, target_weight)
# compute gradient and do update step
optimizer.zero_grad()
loss.backward()
optimizer.step()
to
if isinstance(input, list):
input = input[0].cuda(non_blocking=True)
target = target[0].cuda(non_blocking=True)
target_hm = target
target_weight = target_weight[0].cuda(non_blocking=True)
meta = meta[0]
outputs = model(input)
# target = target[0].cuda(non_blocking=True)
# target_hm = target
# target_weight = target_weight.cuda(non_blocking=True)
loss = criterion(outputs, target, target_weight)
# compute gradient and do update step
optimizer.zero_grad()
loss.backward()
optimizer.step()
because the first element of inputs is the clean one.
Am I correct?
Thanks!
Dear Authors,
Do you mind releasing robust models on few backbones.
Thanks
--kd
argument may have some conflicts. Should we delete this argument?
Thanks!
Here is the test part of res50_256x256_d256x3_adam_lr1e-3_advmix.yaml
. Why is there a key called COCO_BBOX_FILE
? Is there something wrong with this part?
Thanks!
TEST:
BATCH_SIZE_PER_GPU: 32
COCO_BBOX_FILE: 'data/coco/person_detection_results/COCO_val2017_detections_AP_H_56_person.json'
BBOX_THRE: 1.0
IMAGE_THRE: 0.0
IN_VIS_THRE: 0.2
MODEL_FILE: ''
NMS_THRE: 1.0
OKS_THRE: 0.9
FLIP_TEST: true
POST_PROCESS: true
SHIFT_HEATMAP: true
USE_GT_BBOX: true
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.