Skip to content

Commit

Permalink
Add files via upload
Browse files Browse the repository at this point in the history
  • Loading branch information
@TommyZihao
TommyZihao authored Apr 4, 2023
1 parent 57a2fee commit e874eb7
Show file tree
Hide file tree
Showing 25 changed files with 16,502 additions and 0 deletions.
1,792 changes: 1,792 additions & 0 deletions 2022/【A】安装配置MMPose.ipynb
View file

Large diffs are not rendered by default.

1,012 changes: 1,012 additions & 0 deletions 2022/【B1】2D人体关键点 预训练模型预测-Python API.ipynb
View file

Large diffs are not rendered by default.

388 changes: 388 additions & 0 deletions 2022/【B2】2D人体关键点 预训练模型预测-命令行.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,388 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "63f1dd0a-8685-40fd-b5d7-3218f97ac899",
"metadata": {},
"source": [
"# 2D人体关键点 预训练模型预测-命令行\n",
"\n",
"参考文档：https://github.com/open-mmlab/mmpose/blob/master/demo/docs/2d_human_pose_demo.md\n",
"\n",
"作者：同济子豪兄 2022-06-06\n",
"\n",
"如果报错`CUDA out of memory.`则重启前面几个代码的`kernel`即可。"
]
},
{
"cell_type": "markdown",
"id": "4ff424e2-8d49-4f7c-8e81-24a957d1b7b9",
"metadata": {},
"source": [
"## 进入 MMPose 主目录"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "c24d56aa-5528-4561-bb1c-33bcffe2c852",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"['.git',\n",
" '.circleci',\n",
" '.dev_scripts',\n",
" '.github',\n",
" '.gitignore',\n",
" '.pre-commit-config.yaml',\n",
" '.pylintrc',\n",
" '.readthedocs.yml',\n",
" 'CITATION.cff',\n",
" 'LICENSE',\n",
" 'MANIFEST.in',\n",
" 'README.md',\n",
" 'README_CN.md',\n",
" 'configs',\n",
" 'demo',\n",
" 'docker',\n",
" 'docs',\n",
" 'mmpose',\n",
" 'model-index.yml',\n",
" 'pytest.ini',\n",
" 'requirements.txt',\n",
" 'requirements',\n",
" 'resources',\n",
" 'setup.cfg',\n",
" 'setup.py',\n",
" 'tests',\n",
" 'tools',\n",
" 'mmpose.egg-info',\n",
" 'checkpoints',\n",
" 'outputs',\n",
" 'data']"
]
},
"execution_count": 1,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import os\n",
"os.chdir('mmpose')\n",
"os.listdir()"
]
},
{
"cell_type": "markdown",
"id": "c272e015-6112-495d-b407-531fcf46b9bd",
"metadata": {
"tags": []
},
"source": [
"## 自顶向下`top_down`算法"
]
},
{
"cell_type": "markdown",
"id": "92d4b2a2-cdf0-40c9-99d6-e5c0bdfae73a",
"metadata": {},
"source": [
"### 用目标检测预测框作为`top_down`算法的输入框输入"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "8cddc453-bb4d-49e8-8b70-b3be1154712c",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"load checkpoint from http path: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth\n",
"load checkpoint from http path: https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth\n"
]
}
],
"source": [
"!python demo/top_down_img_demo_with_mmdet.py \\\n",
" demo/mmdetection_cfg/faster_rcnn_r50_fpn_coco.py \\\n",
" https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \\\n",
" configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrnet_w48_coco_256x192.py \\\n",
" https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth \\\n",
" --img data/TongjiDancer.png \\\n",
" --out-img-root outputs/B2/B2_1_img"
]
},
{
"cell_type": "markdown",
"id": "ee67f933-2a6f-4146-88b5-c53d1399e866",
"metadata": {},
"source": [
"### 用标注框作为`top_down`算法的输入框输入"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "cd1ae34f-1fc2-4374-8ad8-c62169c2a4b8",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"loading annotations into memory...\n",
"Done (t=0.00s)\n",
"creating index...\n",
"index created!\n",
"load checkpoint from http path: https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth\n"
]
}
],
"source": [
"!python demo/top_down_img_demo.py \\\n",
" configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrnet_w48_coco_256x192.py \\\n",
" https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth \\\n",
" --img-root tests/data/coco/ \\\n",
" --json-file tests/data/coco/test_coco.json \\\n",
" --out-img-root outputs/B2/B2_2_gt_img"
]
},
{
"cell_type": "markdown",
"id": "ea7600f5-c4be-47f5-95b4-147213edabed",
"metadata": {},
"source": [
"### 单帧输入模型的视频预测"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "e9928a78-d7f2-42b1-a3fa-e64035e4a4f9",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initializing model...\n",
"load checkpoint from http path: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth\n",
"load checkpoint from http path: https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth\n",
"Running inference...\n",
"[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 99/99, 3.9 task/s, elapsed: 25s, ETA: 0s\n"
]
}
],
"source": [
"!python demo/top_down_video_demo_with_mmdet.py \\\n",
" demo/mmdetection_cfg/faster_rcnn_r50_fpn_coco.py \\\n",
" https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \\\n",
" configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrnet_w48_coco_256x192.py \\\n",
" https://download.openmmlab.com/mmpose/top_down/hrnet/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth \\\n",
" --video-path data/mot_people_short.mp4 \\\n",
" --bbox-thr 0.8 \\\n",
" --out-video-root outputs/B2/B2_3_video_single_frame"
]
},
{
"cell_type": "markdown",
"id": "6cd383ab-0135-4a44-9c3d-659f32b02954",
"metadata": {},
"source": [
"### 多帧输入模型的视频预测"
]
},
{
"cell_type": "markdown",
"id": "86c42b28-a6fa-4495-8142-02e0251691e3",
"metadata": {},
"source": [
"使用`--use-multi-frames`参数，将视频前后多帧画面输入模型用于姿态预测。\n",
"\n",
"使用`--online`参数，仅输入该帧之前的帧，不输入该帧之后的帧。"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "01931340-af91-4d33-922c-1cad7888865c",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initializing model...\n",
"load checkpoint from http path: https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth\n",
"load checkpoint from http path: https://download.openmmlab.com/mmpose/top_down/posewarper/hrnet_w48_posetrack18_384x288_posewarper_stage2-4abf88db_20211130.pth\n",
"Running inference...\n",
"[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 99/99, 0.5 task/s, elapsed: 183s, ETA: 0s\n"
]
}
],
"source": [
"!python demo/top_down_video_demo_with_mmdet.py \\\n",
" demo/mmdetection_cfg/faster_rcnn_r50_fpn_coco.py \\\n",
" https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \\\n",
" configs/body/2d_kpt_sview_rgb_vid/posewarper/posetrack18/hrnet_w48_posetrack18_384x288_posewarper_stage2.py \\\n",
" https://download.openmmlab.com/mmpose/top_down/posewarper/hrnet_w48_posetrack18_384x288_posewarper_stage2-4abf88db_20211130.pth \\\n",
" --video-path data/mot_people_short.mp4 \\\n",
" --out-video-root outputs/B2/B2_4_multi_frames \\\n",
" --use-multi-frames --online"
]
},
{
"cell_type": "markdown",
"id": "8e3d85ba-ad01-4c1b-8e15-07642829e028",
"metadata": {},
"source": [
"### 全图输入模型的视频预测\n",
"\n",
"不提取人体检测框，直接将全图输入至姿态估计模型中。\n",
"\n",
"仅适用于视频中人体始终在画面中央的场景。\n",
"\n",
"仅适用于单人。\n",
"\n",
"扩展阅读：Mediapipe Blaze Pose单人实时人体姿态估计：https://www.bilibili.com/video/BV1dL4y1h7Q6"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "b698955a-ef57-41be-aa6d-3effacb90f3b",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Initializing model...\n",
"load checkpoint from http path: https://download.openmmlab.com/mmpose/top_down/vipnas/vipnas_res50_coco_256x192-cc43b466_20210624.pth\n",
"Downloading: \"https://download.openmmlab.com/mmpose/top_down/vipnas/vipnas_res50_coco_256x192-cc43b466_20210624.pth\" to /home/featurize/.cache/torch/hub/checkpoints/vipnas_res50_coco_256x192-cc43b466_20210624.pth\n",
"100%|██████████████████████████████████████| 28.0M/28.0M [00:00<00:00, 89.6MB/s]\n",
"Running inference...\n",
"[>>>>>>>>>>>>>>>>>>>>>>>>> ] 645/729, 17.3 task/s, elapsed: 37s, ETA: 5s\n"
]
}
],
"source": [
"!python demo/top_down_video_demo_full_frame_without_det.py \\\n",
" configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/vipnas_res50_coco_256x192.py \\\n",
" https://download.openmmlab.com/mmpose/top_down/vipnas/vipnas_res50_coco_256x192-cc43b466_20210624.pth \\\n",
" --video-path data/solo_dance.mp4 \\\n",
" --out-video-root outputs/B2/B2_5_full_img"
]
},
{
"cell_type": "markdown",
"id": "dd36abfc-31f5-4e52-866d-87e9e670e718",
"metadata": {},
"source": [
"## 自底向上`Bottom-Up`算法"
]
},
{
"cell_type": "markdown",
"id": "5078404b-c30d-49f8-bb97-a32045026523",
"metadata": {},
"source": [
"### 单张图像预测"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "d434a232-411c-44f3-ae64-f7117b8dcade",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"load checkpoint from http path: https://download.openmmlab.com/mmpose/bottom_up/hrnet_w32_coco_512x512-bcb8c247_20200816.pth\n",
"Downloading: \"https://download.openmmlab.com/mmpose/bottom_up/hrnet_w32_coco_512x512-bcb8c247_20200816.pth\" to /home/featurize/.cache/torch/hub/checkpoints/hrnet_w32_coco_512x512-bcb8c247_20200816.pth\n",
"100%|█████████████████████████████████████████| 109M/109M [00:01<00:00, 104MB/s]\n",
"[ ] 0/1, elapsed: 0s, ETA:/home/featurize/work/MMPose教程/mmpose/mmpose/core/post_processing/group.py:240: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').\n",
" y = ind // W\n",
"[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>] 1/1, 0.1 task/s, elapsed: 13s, ETA: 0s\n"
]
}
],
"source": [
"!python demo/bottom_up_img_demo.py \\\n",
" configs/body/2d_kpt_sview_rgb_img/associative_embedding/coco/hrnet_w32_coco_512x512.py \\\n",
" https://download.openmmlab.com/mmpose/bottom_up/hrnet_w32_coco_512x512-bcb8c247_20200816.pth \\\n",
" --img data/TongjiDancer.png \\\n",
" --out-img-root outputs/B2/B2_6_bottom_up_img"
]
},
{
"cell_type": "markdown",
"id": "8bbca5c2-d7b3-4e93-b1f1-5b3fadcb0c2c",
"metadata": {},
"source": [
"### 视频预测"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "d5eba779-3a45-4c11-baea-ced185ab07ff",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"load checkpoint from http path: https://download.openmmlab.com/mmpose/bottom_up/hrnet_w32_coco_512x512-bcb8c247_20200816.pth\n",
"/home/featurize/work/MMPose教程/mmpose/mmpose/core/post_processing/group.py:240: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').\n",
" y = ind // W\n"
]
}
],
"source": [
"!python demo/bottom_up_video_demo.py \\\n",
" configs/body/2d_kpt_sview_rgb_img/associative_embedding/coco/hrnet_w32_coco_512x512.py \\\n",
" https://download.openmmlab.com/mmpose/bottom_up/hrnet_w32_coco_512x512-bcb8c247_20200816.pth \\\n",
" --video-path data/mot_people_short.mp4 \\\n",
" --out-video-root outputs/B2/B2_7_bottom_up_video"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "d4b76e65-1fe0-4bac-b715-063d52c22895",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.7.10"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Loading

0 comments on commit e874eb7

Please sign in to comment.