I am trying to run your code using a webcam.
# $ python3 demos/video.py --weights runs/DirectMHP/agora_m_1280_e300_t40_lw010/weights/best.pt --data data/agora_coco.yaml --device 0 --conf-thres 0.3 --start 0 --thickness 3
import sys
from pathlib import Path
FILE = Path(__file__).absolute()
sys.path.append(FILE.parents[1].as_posix())
import argparse
import torch
import cv2
import yaml
import imageio
from tqdm import tqdm
import os.path as osp
import numpy as np
from utils.torch_utils import select_device, time_sync
from utils.general import check_img_size, scale_coords, non_max_suppression
from utils.datasets import LoadImages
from utils.plots import plot_3axis_Zaxis
from models.experimental import attempt_load
if __name__ == '__main__':
parser = argparse.ArgumentParser()
# video options
# parser.add_argument('-p', '--video-path', default='', help='path to video file')
parser.add_argument('--cam',
dest='cam_id', help='Camera device id to use [0]',
default=0, type=int)
parser.add_argument('--data', type=str,
default='/home/redhwan/2/HPE/DirectMHP/data/agora_coco.yaml'
# default = 'data/agora_coco.yaml'
)
parser.add_argument('--imgsz', type=int, default=1280)
parser.add_argument('--save-size', type=int, default=1080)
parser.add_argument('--weights', default='/home/redhwan/2/HPE/DirectMHP/runs/DirectMHP/agora_m_1280_e300_t40_lw010/weights/best.pt')
parser.add_argument('--device', default=0, help='cuda device, i.e. 0 or cpu')
parser.add_argument('--conf-thres', type=float, default=0.7, help='confidence threshold')
parser.add_argument('--iou-thres', type=float, default=0.45, help='NMS IoU threshold')
parser.add_argument('--scales', type=float, nargs='+', default=[1])
parser.add_argument('--start', type=int, default=0, help='start time (s)')
parser.add_argument('--end', type=int, default=-1, help='end time (s), -1 for remainder of video')
parser.add_argument('--color', type=int, nargs='+', default=[255, 255, 255], help='head bbox color')
parser.add_argument('--thickness', type=int, default=2, help='thickness of Euler angle lines')
parser.add_argument('--alpha', type=float, default=0.4, help='head bbox and head pose alpha')
parser.add_argument('--display', action='store_true', help='display inference results')
parser.add_argument('--fps-size', type=int, default=1)
parser.add_argument('--gif', action='store_true', help='create gif')
parser.add_argument('--gif-size', type=int, nargs='+', default=[480, 270])
args = parser.parse_args()
with open(args.data) as f:
data = yaml.safe_load(f) # load data dict
device = select_device(args.device, batch_size=1)
print('Using device: {}'.format(device))
model = attempt_load(args.weights, map_location=device) # load FP32 model
# stride = int(model.stride.max()) # model stride
# imgsz = check_img_size(args.imgsz, s=stride) # check image size
# if device.type != 'cpu':
# # model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters()))) # run once
# model(torch.zeros(1, 3, imgsz, imgsz).to(device)) # run once
# frames = []
cap = cv2.VideoCapture(0)
while True:
ret, img = cap.read()
# if ret:
# frames.append(img)
# dataset = LoadImages(frame, img_size=imgsz, stride=stride, auto=True)
# video = np.stack(frames, axis=0) # dimensions (T, H, W, C)
# dataset = LoadImages(video, img_size=imgsz, stride=stride, auto=True)
# if device.type != 'cpu':
# model(torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters()))) # run once
# cap = dataset.cap
# cap.set(cv2.CAP_PROP_POS_MSEC, args.start * 1000)
# fps = cap.get(cv2.CAP_PROP_FPS)
# if args.end == -1:
# n = int(cap.get(cv2.CAP_PROP_FRAME_COUNT) - fps * args.start)
# else:
# n = int(fps * (args.end - args.start))
# h = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
# w = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
# gif_frames = []
# out_path = '{}_{}'.format(osp.splitext(args.video_path)[0], "DirectMHP")
# out_path = '{}_{}'.format(osp.splitext(args.video_path)[0], "DirectMHP")
# print("fps:", fps, "\t total frames:", n, "\t out_path:", out_path)
# write_video = not args.display and not args.gif
# if write_video:
# # writer = cv2.VideoWriter(out_path + '.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps, (w, h))
# writer = cv2.VideoWriter(out_path + '.mp4', cv2.VideoWriter_fourcc(*'mp4v'), fps,
# (int(args.save_size * w / h), args.save_size))
# dataset = tqdm(dataset, desc='Running inference', total=n)
# t0 = time_sync()
# for i, (path, img, im0, _) in enumerate(dataset):
print(len(img.shape))
img = torch.from_numpy(img).to(device)
# img = img / 255.0 # 0 - 255 to 0.0 - 1.0
# if len(img.shape) == 3:
# img = img[None] # expand for batch dim
out_ori = model(img, augment=True, scales=args.scales)[0]
# out_ori = model(img, augment=True, scales=args.scales)[1]
# out_ori = model(img, size=321)
out = non_max_suppression(out_ori, args.conf_thres, args.iou_thres, num_angles=data['num_angles'])
# predictions (Array[N, 9]), x1, y1, x2, y2, conf, class, pitch, yaw, roll
bboxes = scale_coords(img.shape[2:], out[0][:, :4], im0.shape[:2]).cpu().numpy() # native-space pred
scores = out[0][:, 4].cpu().numpy()
pitchs_yaws_rolls = out[0][:, 6:].cpu().numpy() # N*3
im0_copy = im0.copy()
# draw head bboxes and pose
for j, [x1, y1, x2, y2] in enumerate(bboxes):
im0_copy = cv2.rectangle(im0_copy, (int(x1), int(y1)), (int(x2), int(y2)),
args.color, thickness=args.thickness)
# im0_copy = cv2.putText(im0_copy, str(round(scores[j], 3)), (int(x1), int(y1)),
# cv2.FONT_HERSHEY_PLAIN, 0.7, (255,255,255), thickness=2)
pitch = (pitchs_yaws_rolls[j][0] - 0.5) * 180
yaw = (pitchs_yaws_rolls[j][1] - 0.5) * 360
roll = (pitchs_yaws_rolls[j][2] - 0.5) * 180
im0_copy = plot_3axis_Zaxis(im0_copy, yaw, pitch, roll, tdx=(x1 + x2) / 2, tdy=(y1 + y2) / 2,
size=max(y2 - y1, x2 - x1) * 0.8, thickness=args.thickness)
im0 = cv2.addWeighted(im0, args.alpha, im0_copy, 1 - args.alpha, gamma=0)
if i == 0:
t = time_sync() - t0
else:
t = time_sync() - t1
if not args.gif and args.fps_size:
cv2.putText(im0, '{:.1f} FPS'.format(1 / t), (5 * args.fps_size, 25 * args.fps_size),
cv2.FONT_HERSHEY_SIMPLEX, args.fps_size, (255, 255, 255), thickness=2 * args.fps_size)
if args.gif:
gif_img = cv2.cvtColor(cv2.resize(im0, dsize=tuple(args.gif_size)), cv2.COLOR_RGB2BGR)
if args.fps_size:
cv2.putText(gif_img, '{:.1f} FPS'.format(1 / t), (5 * args.fps_size, 25 * args.fps_size),
cv2.FONT_HERSHEY_SIMPLEX, args.fps_size, (255, 255, 255), thickness=2 * args.fps_size)
gif_frames.append(gif_img)
elif write_video:
im0 = cv2.resize(im0, dsize=(int(args.save_size * w / h), args.save_size))
writer.write(im0)
else:
cv2.imshow('', im0)
cv2.waitKey(1)
t1 = time_sync()
if i == n - 1:
break
cv2.destroyAllWindows()
cap.release()
if write_video:
writer.release()
if args.gif:
print('Saving GIF...')
with imageio.get_writer(out_path + '.gif', mode="I", fps=fps) as writer:
for idx, frame in tqdm(enumerate(gif_frames)):
writer.append_data(frame)
/usr/bin/python3.8 /home/redhwan/2/HPE/DirectMHP/demos/demo_Redhwan.py
Using device: cuda:0
3
Traceback (most recent call last):
File "/home/redhwan/2/HPE/DirectMHP/demos/demo_Redhwan.py", line 111, in <module>
out_ori = model(img, augment=True, scales=args.scales)[0]
File "/home/redhwan/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/redhwan/2/HPE/DirectMHP/models/yolo.py", line 131, in forward
return self.forward_augment(x, s=scales, f=flips) # augmented inference, None
File "/home/redhwan/2/HPE/DirectMHP/models/yolo.py", line 142, in forward_augment
yi, train_out_i = self.forward_once(xi) # forward
File "/home/redhwan/2/HPE/DirectMHP/models/yolo.py", line 167, in forward_once
x = m(x) # run
File "/home/redhwan/.local/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/redhwan/2/HPE/DirectMHP/models/common.py", line 206, in forward
return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1))
RuntimeError: Sizes of tensors must match except in dimension 2. Got 1 and 2 (The offending index is 0)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent