库名称: transformers
许可证: apache-2.0
基础模型:

• usyd-community/vitpose-plus-huge
  管道标签: 关键点检测

SynthPose (Transformers 🤗 VitPose 巨型变体)

SynthPose模型由Yoni Gozlan、Antoine Falisse、Scott Uhlrich、Anthony Gatti、Michael Black和Akshay Chaudhari在论文OpenCapBench: 连接姿态估计与生物力学的基准中提出。

该模型由Yoni Gozlan贡献。

预期用途

该模型采用VitPose巨型主干网络。
SynthPose是一种新方法，通过使用合成数据，能够对预训练的2D人体姿态模型进行微调，以预测任意更密集的关键点集，从而实现精确的运动学分析。
更多细节请参阅OpenCapBench: 连接姿态估计与生物力学的基准。
此特定变体在运动捕捉设置中常见的关键点集上进行了微调，并包含COCO关键点。

模型预测以下52个标记点：

{
    0: "鼻子",
    1: "左眼",
    2: "右眼",
    3: "左耳",
    4: "右耳",
    5: "左肩",
    6: "右肩",
    7: "左肘",
    8: "右肘",
    9: "左腕",
    10: "右腕",
    11: "左髋",
    12: "右髋",
    13: "左膝",
    14: "右膝",
    15: "左踝",
    16: "右踝",
    17: "胸骨",
    18: "右肩",
    19: "左肩",
    20: "右外肘",
    21: "左外肘",
    22: "右中肘",
    23: "左中肘",
    24: "右外腕",
    25: "左外腕",
    26: "右中腕",
    27: "左中腕",
    28: "右髂前上棘",
    29: "左髂前上棘",
    30: "右髂后上棘",
    31: "左髂后上棘",
    32: "右膝",
    33: "左膝",
    34: "右中膝",
    35: "左中膝",
    36: "右踝",
    37: "左踝",
    38: "右中踝",
    39: "左中踝",
    40: "右第五跖骨",
    41: "左第五跖骨",
    42: "右趾",
    43: "左趾",
    44: "右大脚趾",
    45: "左大脚趾",
    46: "左跟骨",
    47: "右跟骨",
    48: "第七颈椎",
    49: "第二腰椎",
    50: "第十一胸椎",
    51: "第六胸椎",
}

其中前17个关键点为COCO关键点，后35个为解剖学标记点。

使用方法

图像推理

以下是如何加载模型并在图像上运行推理的示例：

import torch
import requests
import numpy as np

from PIL import Image

from transformers import (
    AutoProcessor,
    RTDetrForObjectDetection,
    VitPoseForPoseEstimation,
)

device = "cuda" if torch.cuda.is_available() else "cpu"

url = "http://farm4.staticflickr.com/3300/3416216247_f9c6dfc939_z.jpg"
image = Image.open(requests.get(url, stream=True).raw)

# ------------------------------------------------------------------------
# 阶段1. 检测图像中的人体
# ------------------------------------------------------------------------

# 可根据需要选择检测器
person_image_processor = AutoProcessor.from_pretrained("PekingU/rtdetr_r50vd_coco_o365")
person_model = RTDetrForObjectDetection.from_pretrained("PekingU/rtdetr_r50vd_coco_o365", device_map=device)

inputs = person_image_processor(images=image, return_tensors="pt").to(device)

with torch.no_grad():
    outputs = person_model(**inputs)

results = person_image_processor.post_process_object_detection(
    outputs, target_sizes=torch.tensor([(image.height, image.width)]), threshold=0.3
)
result = results[0]  # 取第一张图像的结果

# COCO数据集中人体标签为0
person_boxes = result["boxes"][result["labels"] == 0]
person_boxes = person_boxes.cpu().numpy()

# 将框从VOC格式(x1, y1, x2, y2)转换为COCO格式(x1, y1, w, h)
person_boxes[:, 2] = person_boxes[:, 2] - person_boxes[:, 0]
person_boxes[:, 3] = person_boxes[:, 3] - person_boxes[:, 1]

# ------------------------------------------------------------------------
# 阶段2. 为检测到的每个人体预测关键点
# ------------------------------------------------------------------------

image_processor = AutoProcessor.from_pretrained("yonigozlan/synthpose-vitpose-huge-hf")
model = VitPoseForPoseEstimation.from_pretrained("yonigozlan/synthpose-vitpose-huge-hf", device_map=device)

inputs = image_processor(image, boxes=[person_boxes], return_tensors="pt").to(device)

with torch.no_grad():
    outputs = model(**inputs)

pose_results = image_processor.post_process_pose_estimation(outputs, boxes=[person_boxes])
image_pose_result = pose_results[0]  # 第一张图像的结果

使用supervision库可视化

import supervision as sv

xy = torch.stack([pose_result['keypoints'] for pose_result in image_pose_result]).cpu().numpy()
scores = torch.stack([pose_result['scores'] for pose_result in image_pose_result]).cpu().numpy()

key_points = sv.KeyPoints(
    xy=xy, confidence=scores
)

vertex_annotator = sv.VertexAnnotator(
    color=sv.Color.PINK,
    radius=2
)

annotated_frame = vertex_annotator.annotate(
    scene=image.copy(),
    key_points=key_points
)
annotated_frame

高级手动可视化

import math
import cv2

def draw_points(image, keypoints, scores, pose_keypoint_color, keypoint_score_threshold, radius, show_keypoint_weight):
    if pose_keypoint_color is not None:
        assert len(pose_keypoint_color) == len(keypoints)
    for kid, (kpt, kpt_score) in enumerate(zip(keypoints, scores)):
        x_coord, y_coord = int(kpt[0]), int(kpt[1])
        if kpt_score > keypoint_score_threshold:
            color = tuple(int(c) for c in pose_keypoint_color[kid])
            if show_keypoint_weight:
                cv2.circle(image, (int(x_coord), int(y_coord)), radius, color, -1)
                transparency = max(0, min(1, kpt_score))
                cv2.addWeighted(image, transparency, image, 1 - transparency, 0, dst=image)
            else:
                cv2.circle(image, (int(x_coord), int(y_coord)), radius, color, -1)

def draw_links(image, keypoints, scores, keypoint_edges, link_colors, keypoint_score_threshold, thickness, show_keypoint_weight, stick_width = 2):
    height, width, _ = image.shape
    if keypoint_edges is not None and link_colors is not None:
        assert len(link_colors) == len(keypoint_edges)
        for sk_id, sk in enumerate(keypoint_edges):
            x1, y1, score1 = (int(keypoints[sk[0], 0]), int(keypoints[sk[0], 1]), scores[sk[0]])
            x2, y2, score2 = (int(keypoints[sk[1], 0]), int(keypoints[sk[1], 1]), scores[sk[1]])
            if (
                x1 > 0
                and x1 < width
                and y1 > 0
                and y1 < height
                and x2 > 0
                and x2 < width
                and y2 > 0
                and y2 < height
                and score1 > keypoint_score_threshold
                and score2 > keypoint_score_threshold
            ):
                color = tuple(int(c) for c in link_colors[sk_id])
                if show_keypoint_weight:
                    X = (x1, x2)
                    Y = (y1, y2)
                    mean_x = np.mean(X)
                    mean_y = np.mean(Y)
                    length = ((Y[0] - Y[1]) ** 2 + (X[0] - X[1]) ** 2) ** 0.5
                    angle = math.degrees(math.atan2(Y[0] - Y[1], X[0] - X[1]))
                    polygon = cv2.ellipse2Poly(
                        (int(mean_x), int(mean_y)), (int(length / 2), int(stick_width)), int(angle), 0, 360, 1
                    )
                    cv2.fillConvexPoly(image, polygon, color)
                    transparency = max(0, min(1, 0.5 * (keypoints[sk[0], 2] + keypoints[sk[1], 2])))
                    cv2.addWeighted(image, transparency, image, 1 - transparency, 0, dst=image)
                else:
                    cv2.line(image, (x1, y1), (x2, y2), color, thickness=thickness)


# 注意: keypoint_edges和调色板是数据集特定的
keypoint_edges = model.config.edges

palette = np.array(
    [
        [255, 128, 0],
        [255, 153, 51],
        [255, 178, 102],
        [230, 230, 0],
        [255, 153, 255],
        [153, 204, 255],
        [255, 102, 255],
        [255, 51, 255],
        [102, 178, 255],
        [51, 153, 255],
        [255, 153, 153],
        [255, 102, 102],
        [255, 51, 51],
        [153, 255, 153],
        [102, 255, 102],
        [51, 255, 51],
        [0, 255, 0],
        [0, 0, 255],
        [255, 0, 0],
        [255, 255, 255],
    ]
)

link_colors = palette[[0, 0, 0, 0, 7, 7, 7, 9, 9, 9, 9, 9, 16, 16, 16, 16, 16, 16, 16]]
keypoint_colors = palette[[16, 16, 16, 16, 16, 9, 9, 9, 9, 9, 9, 0, 0, 0, 0, 0, 0]+[4]*(52-17)]

numpy_image = np.array(image)

for pose_result in image_pose_result:
    scores = np.array(pose_result["scores"])
    keypoints = np.array(pose_result["keypoints"])

    # 在图像上绘制每个点
    draw_points(numpy_image, keypoints, scores, keypoint_colors, keypoint_score_threshold=0.3, radius=2, show_keypoint_weight=False)

    # 绘制连接线
    draw_links(numpy_image, keypoints, scores, keypoint_edges, link_colors, keypoint_score_threshold=0.3, thickness=1, show_keypoint_weight=False)

pose_image = Image.fromarray(numpy_image)
pose_image