【全网首发】2021年中国高校大数据挑战赛B题完整代码版本二

效果

完整代码&＃xff1a;&＃xff08;太多了&＃xff0c;我就不贴全部&＃xff0c;需要的加群&＃xff1a;912166339&＃xff09;
贴一部分&＃xff1a;

# -*- coding:utf-8 -*-
import cv2
import time
import argparse
import numpy as np
from PIL import Image
from utils.anchor_generator import generate_anchors
from utils.anchor_decode import decode_bbox
from utils.nms import single_class_non_max_suppression
from load_model.pytorch_loader import load_pytorch_model, pytorch_inference
model &＃61; load_pytorch_model(&＃39;models/face_mask_detection.pth&＃39;);
# anchor configuration
feature_map_sizes &＃61; [[33, 33], [17, 17], [9, 9], [5, 5], [3, 3]]
anchor_sizes &＃61; [[0.04, 0.056], [0.08, 0.11], [0.16, 0.22], [0.32, 0.45], [0.64, 0.72]]
anchor_ratios &＃61; [[1, 0.62, 0.42]] * 5
# generate anchors
anchors &＃61; generate_anchors(feature_map_sizes, anchor_sizes, anchor_ratios)
# for inference , the batch size is 1, the model output shape is [1, N, 4],
# so we expand dim for anchors to [1, anchor_num, 4]
anchors_exp &＃61; np.expand_dims(anchors, axis&＃61;0)
id2class &＃61; {
0: &＃39;Mask&＃39;, 1: &＃39;NoMask&＃39;}
def inference(image,
conf_thresh&＃61;0.5,
iou_thresh&＃61;0.4,
target_shape&＃61;(160, 160),
draw_result&＃61;True,
show_result&＃61;True
):
&＃39;&＃39;&＃39;
Main function of detection inference
:param image: 3D numpy array of image
:param conf_thresh: the min threshold of classification probabity.
:param iou_thresh: the IOU threshold of NMS
:param target_shape: the model input size.
:param draw_result: whether to daw bounding box to the image.
:param show_result: whether to display the image.
:return:
&＃39;&＃39;&＃39;
# image &＃61; np.copy(image)
output_info &＃61; []
height, width, _ &＃61; image.shape
image_resized &＃61; cv2.resize(image, target_shape)
image_np &＃61; image_resized / 255.0 # 归一化到0~1
image_exp &＃61; np.expand_dims(image_np, axis&＃61;0)
image_transposed &＃61; image_exp.transpose((0, 3, 1, 2))
y_bboxes_output, y_cls_output &＃61; pytorch_inference(model, image_transposed)
# remove the batch dimension, for batch is always 1 for inference.
y_bboxes &＃61; decode_bbox(anchors_exp, y_bboxes_output)[0]
y_cls &＃61; y_cls_output[0]
# To speed up, do single class NMS, not multiple classes NMS.
bbox_max_scores &＃61; np.max(y_cls, axis&＃61;1)
bbox_max_score_classes &＃61; np.argmax(y_cls, axis&＃61;1)
# keep_idx is the alive bounding box after nms.
keep_idxs &＃61; single_class_non_max_suppression(y_bboxes,
bbox_max_scores,
conf_thresh&＃61;conf_thresh,
iou_thresh&＃61;iou_thresh,
)
for idx in keep_idxs:
conf &＃61; float(bbox_max_scores[idx])
class_id &＃61; bbox_max_score_classes[idx]
bbox &＃61; y_bboxes[idx]
# clip the coordinate, avoid the value exceed the image boundary.
xmin &＃61; max(0, int(bbox[0] * width))
ymin &＃61; max(0, int(bbox[1] * height))
xmax &＃61; min(int(bbox[2] * width), width)
ymax &＃61; min(int(bbox[3] * height), height)
if draw_result:
if class_id &＃61;&＃61; 0:
color &＃61; (0, 255, 0)
else:
color &＃61; (255, 0, 0)
cv2.rectangle(image, (xmin, ymin), (xmax, ymax), color, 2)
cv2.putText(image, "%s: %.2f" % (id2class[class_id], conf), (xmin &＃43; 2, ymin - 2),
cv2.FONT_HERSHEY_SIMPLEX, 0.8, color)
output_info.append([class_id, conf, xmin, ymin, xmax, ymax])
if show_result:
Image.fromarray(image).show()
return output_info
def run_on_video(video_path, output_video_name, conf_thresh):
cap &＃61; cv2.VideoCapture(video_path)
height &＃61; cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
width &＃61; cap.get(cv2.CAP_PROP_FRAME_WIDTH)
fps &＃61; cap.get(cv2.CAP_PROP_FPS)
fourcc &＃61; cv2.VideoWriter_fourcc(*&＃39;XVID&＃39;)
# writer &＃61; cv2.VideoWriter(output_video_name, fourcc, int(fps), (int(width), int(height)))
total_frames &＃61; cap.get(cv2.CAP_PROP_FRAME_COUNT)
if not cap.isOpened():
raise ValueError("Video open failed.")
return
status &＃61; True
idx &＃61; 0
while status:
start_stamp &＃61; time.time()
status, img_raw &＃61; cap.read()
img_raw &＃61; cv2.cvtColor(img_raw, cv2.COLOR_BGR2RGB)
read_frame_stamp &＃61; time.time()
if (status):
inference(img_raw,
conf_thresh,
iou_thresh&＃61;0.5,
target_shape&＃61;(260, 260),
draw_result&＃61;True,
show_result&＃61;False)
cv2.imshow(&＃39;image&＃39;, img_raw[:, :, ::-1])
cv2.waitKey(1)
inference_stamp &＃61; time.time()
# writer.write(img_raw)
write_frame_stamp &＃61; time.time()
idx &＃43;&＃61; 1
print("%d of %d" % (idx, total_frames))
print("read_frame:%f, infer time:%f, write time:%f" % (read_frame_stamp - start_stamp,
inference_stamp - read_frame_stamp,
write_frame_stamp - inference_stamp))
# writer.release()
if __name__ &＃61;&＃61; "__main__":
parser &＃61; argparse.ArgumentParser(description&＃61;"Face Mask Detection")
parser.add_argument(&＃39;--img-mode&＃39;, type&＃61;int, default&＃61;1, help&＃61;&＃39;set 1 to run on image, 0 to run on video.&＃39;)
parser.add_argument(&＃39;--img-path&＃39;, type&＃61;str, help&＃61;&＃39;path to your image.&＃39;)
parser.add_argument(&＃39;--video-path&＃39;, type&＃61;str, default&＃61;&＃39;0&＃39;, help&＃61;&＃39;path to your video, &＃96;0&＃96; means to use camera.&＃39;)
# parser.add_argument(&＃39;--hdf5&＃39;, type&＃61;str, help&＃61;&＃39;keras hdf5 file&＃39;)
args &＃61; parser.parse_args()
if args.img_mode:
imgPath &＃61; args.img_path
img &＃61; cv2.imread(imgPath)
img &＃61; cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
inference(img, show_result&＃61;True, target_shape&＃61;(260, 260))
else:
video_path &＃61; args.video_path
if args.video_path &＃61;&＃61; &＃39;0&＃39;:
video_path &＃61; 0
run_on_video(video_path, &＃39;&＃39;, conf_thresh&＃61;0.5)


比赛的话&＃xff0c;千叮万嘱&＃xff0c;一定要在我这基础上做修改&＃xff01;&＃xff01;我就不帮大家修改了&＃xff0c;自行发挥&＃xff01;
允许转发本文&＃xff0c;请不要删除本文内容&＃xff0c;谢谢。
更多比赛讨论请加qq群&＃xff1a;912166339&＃xff0c;伸手党勿扰。
欢迎关注个人公众号&＃xff1a;川川菜鸟
CSDN&＃xff1a;川川菜鸟