文件最后提交记录最后更新时间
【fix】完善小模型Atlas 300I DUO硬件描述 Co-authored-by: Niushiya<niushiya1@huawei.com> # message auto-generated for no-merge-commit merge: !7587 merge master into master 【fix】完善小模型Atlas 300I DUO硬件描述 Created-by: niushiya Commit-by: Niushiya Merged-by: ascend-robot Description: ## Motivation 1、完善小模型Atlas 300I DUO硬件描述,补充单芯字段; ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!75875 天前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7537 merge fix_chinese_clip into master [fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 解决Chinese Clip 数据集、模型、方案失效问题 GenPose2 导出逻辑bug修复 ## Modification 更新Chinese Clip数据集下载地址、模型下载地址 更新了依赖项 更新Chinese Clip 的 commit id 为最新版本。 提供了 导出onnx、om的脚本 更新了推理的补丁 修复GenPose2 集体导出onnx 和 om 的检测逻辑bug ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!753727 天前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7537 merge fix_chinese_clip into master [fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 解决Chinese Clip 数据集、模型、方案失效问题 GenPose2 导出逻辑bug修复 ## Modification 更新Chinese Clip数据集下载地址、模型下载地址 更新了依赖项 更新Chinese Clip 的 commit id 为最新版本。 提供了 导出onnx、om的脚本 更新了推理的补丁 修复GenPose2 集体导出onnx 和 om 的检测逻辑bug ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!753727 天前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[feature] genpose2更新om推理路线;支持tracking推理 Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7529 merge genpose2_om into master [feature] genpose2更新om推理路线;支持tracking推理 Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 更新genpose2,支持om推理;支持tracking推理 ## Modification 1.提供了六个分模型转onnx及转om的四个脚本 2.拆分了原网络过于耦合的架构,使得导出onnx成为可能,并保留了pt推理的能力。 3.优化了算子计算的方式 4.打通了infer.py推理的通路。 5.打通了evaluation_tracking.py的通路。 ## Self-test (Optional) 自测显示精度无明显变化,性能有较大提升 ~3000 ms/sample ---> ~650 ms/sample ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] 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!75291 个月前
[fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Co-authored-by: frost_mourne<suyuxuan3@h-partners.com> # message auto-generated for no-merge-commit merge: !7537 merge fix_chinese_clip into master [fix] Chinese Clip 数据集、模型、方案失效。[fix] 小修GenPose2 导出逻辑bug Created-by: frost_mourne Commit-by: frost_mourne Merged-by: ascend-robot Description: ## Motivation 解决Chinese Clip 数据集、模型、方案失效问题 GenPose2 导出逻辑bug修复 ## Modification 更新Chinese Clip数据集下载地址、模型下载地址 更新了依赖项 更新Chinese Clip 的 commit id 为最新版本。 提供了 导出onnx、om的脚本 更新了推理的补丁 修复GenPose2 集体导出onnx 和 om 的检测逻辑bug ## 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**: - [x] The new code needs to comply with the Clean Code specification. - [x] The PR content is self-checked, and the expression can be clear and the writing standardized **After PR**: - [x] CLA has been signed and all committers have signed the CLA in this PR. - [x] The ci-pipeline is passed, Code Check is passed. See merge request: Ascend/ModelZoo-PyTorch!753727 天前
README.md

GenPose++ 推理指导

概述

GenPose++ 采用分段点云和裁剪后的 RGB 图像作为输入,利用 PointNet++ 提取物体的几何特征;同时借助预训练的 2D 基础骨干网络 DINO v2 提取通用语义特征。随后,将这些特征融合,作为扩散模型的条件,生成物体姿态候选及其对应的能量。最后,针对具有非连续对称性(如盒子)的物体,通过聚类解决姿态多模态分布带来的聚集问题,从而有效完成姿态估计。

插件与驱动准备

  • 该模型需要以下插件与驱动
配套 版本 环境准备指导
固件与驱动 24.1.RC3 Pytorch框架推理环境准备
CANN 8.3.RC1 包含kernels包和toolkit包
Python 3.10.14 -
PyTorch 2.1.0 -
Ascend Extension PyTorch 2.1.0.post17 -
说明:支持Atlas 300I DUO/ 310P RC \ \

环境准备

  • 安装conda环境(略)

使用conda创建虚拟环境并激活:

conda create -n genpose2 python==3.10.14
conda activate genpose2
  • 安装推理工具ais_bench

ais_bench安装请参考ais_bench安装指导

获取源码并处理

获取源码,切到指定commit,激活必要环境,安装必要依赖:

# 获取源码
git clone https://gitcode.com/ascend/ModelZoo-PyTorch.git
cd ModelZoo-PyTorch/ACL_PyTorch/built-in/embodied_ai/GenPosePlus
export PYTHONPATH=$PWD:$PYTHONPATH
git clone https://github.com/Omni6DPose/GenPose2.git
cd GenPose2
export PYTHONPATH=$PWD:$PYTHONPATH

# 切换至指定commit
git reset d0993c0

#应用补丁,注意,此命令只能执行一次,再次执行会报错(不影响后续使用)
git apply ../diff.patch
export GPPATH=$PWD

# 激活必要环境
source /usr/local/Ascend/ascend-toolkit/set_env.sh
export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True


#安装相关依赖:
pip3 install -r requirements.txt
pip3 install -r ../requirements.txt

下载配置文件与模型

请按照 Omni6DPoseAPI页面上的说明,下载并整理 Omni6DPose 数据集中的ROPE分类中的000000号数据集,并在项目目录中组织如下

omni6dpose-000000
└──ROPE
   └── 000000
	  ├── 000000_color.png
	  ├── 000000_depth.exr
	  ├── 000000_mask.exr
	  ├── 000000_mask_sam.npz
	  ├── 000000_meta.json
	  ...
	  └── 000945_meta.json
  • 在数据集下载页面ROPE同级目录找到 Meta/路径下内容并复制至 $GPPATH/configs/ 路径,组织如下:
GenPose2
└──configs
   ├── obj_meta.json
   ├── real_obj_meta.json
   └── config.py
  • 模型可在如下连接找到 checkpoints. 请下载至如下路径 $GPPATH/results 并组织如下:
GenPose2
└──results
   └── ckpts
       ├── ScoreNet
       │   └── scorenet.pth
       ├── EnergyNet
       │   └── energynet.pth
       └── ScaleNet
           └── scalenet.pth

推理与评测

Single模式评测

该模式会将数据集下所有数据视为没有关联,随机每一个数据的初始值。

  • 转换pth模型为onnx模型
# 进入项目目录
cd $GPPATH
# 执行转换脚本,运行需要一些时间
python ../export_all_onnx.py
  • 转换onnx模型为om模型
# 执行转换脚本,运行需要一些时间
python ../export_all_om.py
# 可跟参数 --soc_version,缺省为Ascend310P3

执行如下命令进行评测

bash scripts/eval_single_om.sh

Tracking模式评测

该模式会将数据集下所有数据视为有连续关联,非首数据的初始值会基于上一个数据。

  • 转换pth模型为onnx模型
# 进入项目目录
cd $GPPATH
# 执行转换脚本,运行需要一些时间
python ../export_all_onnx.py --batch_size 4
  • 转换onnx模型为om模型
# 执行转换脚本,运行需要一些时间
python ../export_all_om.py --batch_size 4 --soc_version Ascend310P3
# Atlas 300I DUO 设备的 soc_version 为 Ascend310P3 ,310P RC 设备的 soc_version 为 Ascend310P1 
# --soc_version参数可缺省,默认为Ascend310P3

执行如下命令进行评测

bash scripts/eval_tracking_om.sh

注意: 两种方式,非首次推理必须删除上一次缓存,否则所得结果为上一次缓存。 参考删除缓存命令

rm -rf $GPPATH/results/evaluation_results

模型推理性能精度结果:

芯片 推理方式 batchsize iou_mean 参考性能(ms/sample)
300I DUO(单芯) single 16 0.3073 652.20
300I DUO(单芯) tracking 4 0.3038 458.37
310P RC single 16 0.3073 803.23
310P RC tracking 4 0.3036 488.15