dd6fe347创建于 4月9日历史提交

文件	最后提交记录	最后更新时间
configs	init	4 年前
lib	init	4 年前
run	init	4 年前
test	!4671 【fix】批量修改模型python版本，兼容环境上的python3.8版本 * fix python version	3 年前
.keep	init	4 年前
CODE_OF_CONDUCT.md	init	4 年前
LICENSE	init	4 年前
README.md	!5816 Network address of models to be rectified: 23 Merge pull request !5816 from Yss/network_declaration_23	2 年前
SECURITY.md	init	4 年前
modelzoo_level.txt	init	4 年前
public_address_statement.md	fix link validity Co-authored-by: frozenleaves<914814442@qq.com> # message auto-generated for no-merge-commit merge: !7517 merge master into master fix link validity Created-by: frozenn Commit-by: frozenleaves Merged-by: ascend-robot Description: ## Motivation Please describe the motivation of this PR and the goal you want to achieve through this PR. ## Modification Please briefly describe what modification is made in this PR. ## Self-test (Optional) If modifications to this PR may cause/fix function/accuracy/performance DTSs/issues, a self-inspection record needs to be attached. ## BC-breaking (Optional) If there are compatibility issues, such as dependencies on cann/torch_npu versions, they need to be explained in the PR. ## Checklist Before PR: - [ ] The new code needs to comply with the Clean Code specification. - [ ] The PR content is self-checked, and the expression can be clear and the writing standardized After PR: - [ ] CLA has been signed and all committers have signed the CLA in this PR. - [ ] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!7517	1 个月前
requirements.txt	init	4 年前

VoxelPose

This implements training of voxelpose on the MultiHumanPose's Shelf Dataset, mainly modified from microsoft/voxelpose-pytorch.

VoxelPose Detail

Requirements

Install PyTorch (pytorch.org)
pip install -r requirements.txt and then install the specified version of torch, torchvision and apex
Download the dataset from http://campar.in.tum.de/Chair/MultiHumanPose (Shelf Dataset) and extract them under ${POSE_ROOT}/data/Shelf
Put the panoptic_training_pose.pkl under ${POSE_ROOT}/data

Add Shape

Set AscendProject = /usr/local/Ascend #use real path

Add shape [32, 16, 1, 1, 1], [1, 32, 1, 1, 1], [17, 32, 1, 1, 1] into ${AscendProject}/ascend-toolkit/5.0.2/arm64-linux/opp/op_impl/built-in/ai_core/tbe/impl/fractal_z_3d_2_ncdhw.py
Add shape [262144, 1, 1, 16, 16], [128000, 1, 1, 16, 16] into ${AscendProject}/ascend-toolkit/5.0.2/arm64-linux/opp/op_impl/built-in/ai_core/tbe/impl/fractal_z_3d_2_ncdhw.py

Training

To train a model, run the following scripts with the real path of dataset: PS：There is a large jitter on single card training with bs=1, do not perform single card trainning.

# real_data_path = data/shelf
# training 8p accuracy
bash ./test/train_full_8p.sh --data_path=real_data_path

# training 8p performance
bash ./test/train_performance_8p.sh --data_path=real_data_path

# test 8p accuracy
bash test/train_eval_8p.sh

Log path: ${POSE_ROOT}/output/shelf_synthetic/multi_person_posenet_50/prn64_cpn80x80x20/prn64_cpn80x80x20_{time}_train.log # training detail log

VoxelPose training result

说明：由于模型单卡训练（bs=1）抖动大，loss不收敛，暂不持支单卡训练。

名称	精度	FPS	AMP_Type
NPU-8p	97.10	1.267	O1

Statement

For details about the public address of the code in this repository, you can get from the file public_address_statement.md