# yolov3-mxnet
A yolov3 simple implementation in MXNet, based on version 1.2.0 and cuda 9.0(optional), python3.
Works on Windows and Ubuntu 16.04.
<br>
<br>
**new: hybridized, speed up.**
<br>
**new: train demo.**
<br>
<br>
**Detect Part Completed.**
<br>
=======
## Table of Contents
- [yolov3-mxnet](#yolov3-mxnet)
* [Paper](#paper)
* [Installation](#installation)
* [Detection](#detection)
* [Train](#train)
* [Credit](#credit)
## Paper
### YOLOv3: An Incremental Improvement
_Joseph Redmon, Ali Farhadi_
<br \>
**Abstract**
<br \>
We present some updates to YOLO! We made a bunch
of little design changes to make it better. We also trained
this new network that’s pretty swell. It’s a little bigger than
last time but more accurate. It’s still fast though, don’t
worry. At 320 × 320 YOLOv3 runs in 22 ms at 28.2 mAP,
as accurate as SSD but three times faster. When we look
at the old .5 IOU mAP detection metric YOLOv3 is quite
good. It achieves 57.9 AP50 in 51 ms on a Titan X, compared
to 57.5 AP50 in 198 ms by RetinaNet, similar performance
but 3.8× faster. As always, all the code is online at
https://pjreddie.com/yolo/.
[[Paper]](https://pjreddie.com/media/files/papers/YOLOv3.pdf) [[Original Implementation]](https://github.com/pjreddie/darknet)
## Installation
$ cd 项目
$ sudo pip3 install opencv-python mxnet-cu90==1.2.0
## Detection
put your images in ./images, and
$ python detect.py [--gpu GPU ID]
you will get the results in ./results
or you can detect a video file, this need your opencv is compiled with ffmpeg.
$ python detect.py --video VIDEO_FILE
you will get result.avi in ./results
## Train (for reference only)
The IMAGE_FOLDER should contains two directorys, "train" and "train_label", (or four, "train", "train_label", "val", "val_label") label should be xml file like VOC's format.
PS: You can use voc_label.py in https://pjreddie.com/darknet/yolo/ to get train.txt and val.txt, set path to --train and --val instead of --images.
```
train.py [-h] [--epochs EPOCHS] [--images IMAGES FOLDER]
[--batch_size BATCH_SIZE]
[--params WEIGHTS_PATH] [--classes CLASS_PATH]
[--confidence CONF_THRES] [--nms_thresh NMS_THRES]
[--gpu GPU ID] [--prefix PARAMS FILE NAME PREFIX]
```
没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
收起资源包目录
yolo_基于Mxnet实现的YOLOv3目标检测算法.zip (9个子文件)
yolo_基于Mxnet实现的YOLOv3目标检测算法
utils.py 17KB
data
voc.names 134B
get_weights.sh 52B
coco.names 624B
darknet.py 17KB
detect.py 10KB
images
dog-cycle-car.png 339KB
train.py 11KB
README.md 2KB
共 9 条
- 1
资源评论
DdddJMs__135
- 粉丝: 727
- 资源: 217
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功