detect.py
前言
🎉代码仓库地址：https://github.com/Oneflow-Inc/one-yolov5 欢迎star one-yolov5项目 获取最新的动态。 如果您有问题，欢迎在仓库给我们提出宝贵的意见。🌟🌟🌟 如果对您有帮助，欢迎来给我Star呀😊~
源码解读： detect.py
Run inference with a YOLOv5 model on images, videos, directories, streams
1. 导入需要的包和基本配置
import argparse # python的命令行解析的标准模块 可以让我们直接在命令行中就可以向程序中传入参数并让程序运行
import os # os库是Python标准库,包含几百个函数,常用路径操作、进程管理、环境参数等几类。
import platform # 用platform模块可以判断当前的系统环境
import sys # sys系统模块 包含了与Python解释器和它的环境有关的函数
from pathlib import Path # Path将str转换为Path对象 使字符串路径易于操作的模块
import numpy as np # Numpy是使用C语言实现的一个数据计算库
import oneflow as flow # OneFlow框架
import oneflow.backends.cudnn as cudnn
from models.common import DetectMultiBackend
from utils.dataloaders import IMG_FORMATS, VID_FORMATS, LoadImages, LoadStreams
from utils.general import LOGGER, check_file, check_img_size, check_imshow, check_requirements, colorstr, cv2, increment_path, non_max_suppression, print_args, scale_coords, strip_optimizer, xyxy2xywh
from utils.plots import Annotator, colors, save_one_box
from utils.oneflow_utils import select_device, time_sync
FILE = Path(__file__).resolve()
ROOT = FILE.parents[0] # YOLOv5 root directory
if str(ROOT) not in sys.path:
sys.path.append(str(ROOT)) # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd())) # relative
2. opt参数详解
| 参数 | 解析 | |
|---|---|---|
| weights | model path(s) | 模型的权重地址 |
| source | file/dir/URL/glob, 0 for webcam | 测试数据文件(图片或视频)的保存路径 默认data/images |
| data | (optional) dataset.yaml path | 数据集配置文件路径 |
| imgsz | inference size h,w | 网络输入图片的大小 默认640 |
| conf-thres | confidence threshold | object置信度阈值 默认0.25 |
| iou-thres | NMS IoU threshold | 做nms的iou阈值 默认0.45 |
| max-det | maximum detections per image | 每张图片最大的目标个数 默认1000 |
| device | cuda device, i.e. 0 or 0,1,2,3 or cpu | 设置代码执行的设备 cuda device, i.e. 0 or 0,1,2,3 or cpu |
| view-img | show results | 是否展示预测之后的图片或视频 默认False |
| save-txt | save results to *.txt | 是否将预测的框坐标以txt文件格式保存 默认False 会在runs/detect/expn/labels下生成每张图片预测的txt文件 |
| save-conf | save confidences in --save-txt labels | 是否保存预测每个目标的置信度到预测tx文件中 默认False |
| save-crop | save cropped prediction boxes | 是否需要将预测到的目标从原图中扣出来 剪切好 并保存 会在runs/detect/expn下生成crops文件，将剪切的图片保存在里面 默认False |
| nosave | do not save images/videos | 是否不要保存预测后的图片 默认False 就是默认要保存预测后的图片 |
| classes | filter by class: --classes 0, or --classes 0 2 3 | 在nms中是否是只保留某些特定的类 默认是None 就是所有类只要满足条件都可以保留 |
| agnostic-nms | class-agnostic NMS | 进行nms是否也除去不同类别之间的框 默认False |
| augment | augmented inference | 预测是否也要采用数据增强 TTA |
| visualize | visualize features | 可视化特征 |
| update | update all models | 是否将optimizer从ckpt中删除 更新模型 默认False |
| project | save results to project/name | 当前测试结果放在哪个主文件夹下 默认runs/detect |
| name | save results to project/name | 当前测试结果放在run/detect下的文件名 默认是exp |
| exist-ok | existing project/name ok, do not increment | 是否存在当前文件 默认False 一般是 no exist-ok 连用 所以一般都要重新创建文件夹 |
| line-thickness | bounding box thickness (pixels) | 画框的框框的线宽 默认是 3 |
| hide-labels | hide labels | 画出的框框是否需要隐藏label信息 默认False |
| hide-conf | hide confidences | 画出的框框是否需要隐藏label信息 默认False |
| half | use FP16 half-precision inference | 画出的框框是否需要隐藏conf信息 默认False |
| dnn | use OpenCV DNN for ONNX inference | 是否使用半精度 Float16 推理 可以缩短推理时间 但是默认是False |
3 main函数
def main(opt):
# 检查包是否满足requirements对应txt文件的要求
check_requirements(exclude=('tensorboard', 'thop'))
# 执行run 开始推理
run(**vars(opt))
4 run函数
@flow.no_grad()
def run(
weights=ROOT / "yolov5s", # model path(s)
source=ROOT / "data/images", # file/dir/URL/glob, 0 for webcam
data=ROOT / "data/coco128.yaml", # dataset.yaml path
imgsz=(640, 640), # inference size (height, width)
conf_thres=0.25, # confidence threshold
iou_thres=0.45, # NMS IOU threshold
max_det=1000, # maximum detections per image
device="", # cuda device, i.e. 0 or 0,1,2,3 or cpu
view_img=False, # show results
save_txt=False, # save results to *.txt
save_conf=False, # save confidences in --save-txt labels
save_crop=False, # save cropped prediction boxes
nosave=False, # do not save images/videos
classes=None, # filter by class: --class 0, or --class 0 2 3
agnostic_nms=False, # class-agnostic NMS
augment=False, # augmented inference
visualize=False, # visualize features
update=False, # update all models
project=ROOT / "runs/detect", # save results to project/name
name="exp", # save results to project/name
exist_ok=False, # existing project/name ok, do not increment
line_thickness=3, # bounding box thickness (pixels)
hide_labels=False, # hide labels
hide_conf=False, # hide confidences
half=False, # use FP16 half-precision inference
dnn=False, # use OpenCV DNN for ONNX inference
):
source = str(source)
# 是否保存预测后的图片 默认nosave=False 所以只要传入的文件地址不是以.txt结尾 就都是要保存预测后的图片的
save_img = not nosave and not source.endswith(".txt") # save inference images
# 文件类型
is_file = Path(source).suffix[1:] in (IMG_FORMATS + VID_FORMATS)
# 是否是url网络地址
is_url = source.lower().startswith(("rtsp://", "rtmp://", "http://", "https://"))
# 是否是使用webcam 网页数据 一般是Fasle 因为我们一般是使用图片流LoadImages(可以处理图片/视频流文件)
webcam = source.isnumeric() or source.endswith(".txt") or (is_url and not is_file)
if is_url and is_file:
source = check_file(source) # download
# Directories
save_dir = increment_path(Path(project) / name, exist_ok=exist_ok) # 增量运行
(save_dir / "labels" if save_txt else save_dir).mkdir(parents=True, exist_ok=True) # 创建文件夹用于存储输出结果
# Load model
device = select_device(device) # 获取当前主机可用的设备
model = DetectMultiBackend(weights, device=device, dnn=dnn, data=data, fp16=half)
# stride: 模型最大的下采样率 [8, 16, 32] 所有stride一般为32
# names: 得到数据集的所有类的类名
# of : oneflow模型权重文件
stride, names, of = model.stride, model.names, model.of
# 确保输入图片的尺寸imgsz能整除stride 如果不能则调整为能被整除并返回
imgsz = check_img_size(imgsz, s=stride) # check image size
# Dataloader
if webcam: # 一般不会使用webcam模式从网页中获取数据
view_img = check_imshow()
cudnn.benchmark = True # 设置为True，使用CUDNN以加快恒定图像尺寸推断的速度
dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=of)
bs = len(dataset) # batch_size
else: # 一般是直接从source文件目录下直接读取图片或者视频数据
dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=of)
bs = 1 # batch_size
vid_path, vid_writer = [None] * bs, [None] * bs
# Run inference 正式推理
model.warmup(imgsz=(1 if of else bs, 3, *imgsz)) # warmup
seen, windows, dt = 0, [], [0.0, 0.0, 0.0]
for path, im, im0s, vid_cap, s in dataset:
t1 = time_sync()
im = flow.from_numpy(im).to(device)
im = im.half() if model.fp16 else im.float() # uint8 to fp16/32
im /= 255 # 0 - 255 to 0.0 - 1.0
if len(im.shape) == 3:
im = im[None] # expand for batch dim
t2 = time_sync()
dt[0] += t2 - t1
# Inference
visualize = increment_path(save_dir / Path(path).stem, mkdir=True) if visualize else False
pred = model(im, augment=augment, visualize=visualize)
t3 = time_sync()
dt[1] += t3 - t2
# NMS 去掉detection任务重复的检测框。更多请参阅 https://blog.csdn.net/yql_617540298/article/details/89474226
pred = non_max_suppression(pred, conf_thres, iou_thres, classes, agnostic_nms, max_det=max_det)
dt[2] += time_sync() - t3
# Second-stage classifier (optional)
# pred = utils.general.apply_classifier(pred, classifier_model, im, im0s)
# Process predictions
for i, det in enumerate(pred): # per image 对每张图片进行处理 将pred(相对img_size 640)映射回原图img0 size
seen += 1
if webcam: # batch_size >= 1
p, im0, frame = path[i], im0s[i].copy(), dataset.count
s += f"{i}: "
else:
p, im0, frame = path, im0s.copy(), getattr(dataset, "frame", 0)
p = Path(p) # to Path
save_path = str(save_dir / p.name) # im.jpg
txt_path = str(save_dir / "labels" / p.stem) + ("" if dataset.mode == "image" else f"_{frame}") # im.txt
s += "%gx%g " % im.shape[2:] # print string
gn = flow.tensor(im0.shape)[[1, 0, 1, 0]] # normalization gain whwh
imc = im0.copy() if save_crop else im0 # for save_crop
annotator = Annotator(im0, line_width=line_thickness, example=str(names))
if len(det):
# Rescale boxes from img_size to im0 size
# 将预测信息（相对img_size 640）映射回原图 img0 size
det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()
det = det.detach().cpu().numpy()
# Print results
# 输出信息s + 检测到的各个类别的目标个数
for c in np.unique(det[:, -1]):
n = (det[:, -1] == c).sum() # detections per class
s += f"{n} {names[int(c)]}{'s' * (n > 1)}, " # add to string
# Write results
# 保存预测信息: txt、img0上画框、crop_img
for *xyxy, conf, cls in reversed(det):
# 将每个图片的预测信息分别存入save_dir/labels下的xxx.txt中 每行: class_id+score+xywh
if save_txt: # Write to file
# 将xyxy(左上角 + 右下角)格式转换为xywh(中心的 + 宽高)格式 并除以gn(whwh)做归一化 转为list再保存
xywh = (xyxy2xywh(flow.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist() # normalized xywh
line = (cls, *xywh, conf) if save_conf else (cls, *xywh) # label format
with open(f"{txt_path}.txt", "a") as f:
f.write(("%g " * len(line)).rstrip() % line + "\n")
# 在原图上画框 + 将预测到的目标剪切出来
if save_img or save_crop or view_img: # Add bbox to image
c = int(cls) # integer class
label = None if hide_labels else (names[c] if hide_conf else f"{names[c]} {conf:.2f}")
annotator.box_label(xyxy, label, color=colors(c, True))
if save_crop:# 如果需要就将预测到的目标剪切出来 保存成图片 保存在save_dir/crops下
save_one_box(
xyxy,
imc,
file=save_dir / "crops" / names[c] / f"{p.stem}.jpg",
BGR=True,
)
# Stream results
im0 = annotator.result()
if view_img:
if platform.system() == "Linux" and p not in windows:
windows.append(p)
cv2.namedWindow(str(p), cv2.WINDOW_NORMAL | cv2.WINDOW_KEEPRATIO) # allow window resize (Linux)
cv2.resizeWindow(str(p), im0.shape[1], im0.shape[0])
cv2.imshow(str(p), im0)
cv2.waitKey(1) # 1 millisecond
# Save results (image with detections)
if save_img: # 是否需要保存图片或视频（检测后的图片/视频 里面已经被我们画好了框的） img0
if dataset.mode == "image":
cv2.imwrite(save_path, im0)
else: # 'video' or 'stream'
if vid_path[i] != save_path: # new video
vid_path[i] = save_path
if isinstance(vid_writer[i], cv2.VideoWriter):
vid_writer[i].release() # release previous video writer
if vid_cap: # video
fps = vid_cap.get(cv2.CAP_PROP_FPS)
w = int(vid_cap.get(cv2.CAP_PROP_FRAME_WIDTH))
h = int(vid_cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
else: # stream
fps, w, h = 30, im0.shape[1], im0.shape[0]
save_path = str(Path(save_path).with_suffix(".mp4")) # force *.mp4 suffix on results videos
vid_writer[i] = cv2.VideoWriter(save_path, cv2.VideoWriter_fourcc(*"mp4v"), fps, (w, h))
vid_writer[i].write(im0)
# Print time (inference-only)
LOGGER.info(f"{s}Done. ({t3 - t2:.3f}s)")
# Print results
t = tuple(x / seen * 1e3 for x in dt) # speeds per image
LOGGER.info(f"%.1fms pre-process, %.1fms inference, %.1fms NMS per image at shape {(1, 3, *imgsz)}" % t)
if save_txt or save_img:
s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ""
LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}{s}")
if update:
# strip_optimizer函数将optimizer从ckpt中删除 更新模型
strip_optimizer(weights[0]) # update model (to fix SourceChangeWarning)