preprocess_img_pose.py

'''cropping image and pose esimation before entering PIFuHD'''

import torch
import cv2
import numpy as np
from os import path
import sys
sys.path.append(path.abspath('./lightweight-human-pose-estimation.pytorch'))
from models.with_mobilenet import PoseEstimationWithMobileNet
from modules.keypoints import extract_keypoints, group_keypoints
from modules.load_state import load_state
from modules.pose import Pose, track_poses
import demo
import sys

def get_rect(net, images, height_size):
    net = net.eval()

    stride = 8
    upsample_ratio = 4
    num_keypoints = Pose.num_kpts
    previous_poses = []
    delay = 33
    for image in images:
        print("image:"+image)
        rect_path = image.replace('.%s' % (image.split('.')[-1]), '_rect.txt')
        #print("rect_path"+rect_path)
        img = cv2.imread(image, cv2.IMREAD_COLOR)
        orig_img = img.copy()
        orig_img = img.copy()
        heatmaps, pafs, scale, pad = demo.infer_fast(net, img, height_size, stride, upsample_ratio, cpu=True)

        total_keypoints_num = 0
        all_keypoints_by_type = []
        for kpt_idx in range(num_keypoints):  # 19th for bg
            total_keypoints_num += extract_keypoints(heatmaps[:, :, kpt_idx], all_keypoints_by_type, total_keypoints_num)

        pose_entries, all_keypoints = group_keypoints(all_keypoints_by_type, pafs)
        for kpt_id in range(all_keypoints.shape[0]):
            all_keypoints[kpt_id, 0] = (all_keypoints[kpt_id, 0] * stride / upsample_ratio - pad[1]) / scale
            all_keypoints[kpt_id, 1] = (all_keypoints[kpt_id, 1] * stride / upsample_ratio - pad[0]) / scale
        current_poses = []

        rects = []
        for n in range(len(pose_entries)):
            if len(pose_entries[n]) == 0:
                continue
            pose_keypoints = np.ones((num_keypoints, 2), dtype=np.int32) * -1
            valid_keypoints = []
            for kpt_id in range(num_keypoints):
                if pose_entries[n][kpt_id] != -1.0:  # keypoint was found
                    pose_keypoints[kpt_id, 0] = int(all_keypoints[int(pose_entries[n][kpt_id]), 0])
                    pose_keypoints[kpt_id, 1] = int(all_keypoints[int(pose_entries[n][kpt_id]), 1])
                    valid_keypoints.append([pose_keypoints[kpt_id, 0], pose_keypoints[kpt_id, 1]])
            valid_keypoints = np.array(valid_keypoints)

            if pose_entries[n][10] != -1.0 or pose_entries[n][13] != -1.0:
              pmin = valid_keypoints.min(0)
              pmax = valid_keypoints.max(0)

              center = (0.5 * (pmax[:2] + pmin[:2])).astype(int)
              radius = int(0.65 * max(pmax[0]-pmin[0], pmax[1]-pmin[1]))
            elif pose_entries[n][10] == -1.0 and pose_entries[n][13] == -1.0 and pose_entries[n][8] != -1.0 and pose_entries[n][11] != -1.0:
              # if leg is missing, use pelvis to get cropping
              center = (0.5 * (pose_keypoints[8] + pose_keypoints[11])).astype(np.int)
              radius = int(1.45*np.sqrt(((center[None,:] - valid_keypoints)**2).sum(1)).max(0))
              center[1] += int(0.05*radius)
            else:
              center = np.array([img.shape[1]//2,img.shape[0]//2])
              radius = max(img.shape[1]//2,img.shape[0]//2)

            x1 = center[0] - radius
            y1 = center[1] - radius

            rects.append([x1, y1, 2*radius, 2*radius])

        np.savetxt(rect_path, np.array(rects), fmt='%d')

net = PoseEstimationWithMobileNet()
checkpoint = torch.load('./checkpoints/checkpoint_iter_370000.pth', map_location='cpu')
load_state(net, checkpoint)

#image_path = './pifuhd/sample_images/%s' % filename
image_path = str(sys.argv[1])
#print("Input Image:"+str(image_path))
get_rect(net, [image_path], 512)