인공지능

Implementation of YOLO v3 object detector in TensorFlow

여니두 2019. 8. 22.

https://github.com/YunYang1994/tensorflow-yolov3

 

YunYang1994/tensorflow-yolov3

🔥 pure tensorflow Implement of YOLOv3 with support to train your own dataset - YunYang1994/tensorflow-yolov3

github.com

 

2. Quick start

* error1) 

>> requirements.txt 수정

 

numpy==1.17.0 
Pillow==5.3.0 
scipy==1.1.0 
tensorflow-gpu==1.14.0 
wget==3.2 
seaborn==0.9.0

 

- checkpoint

* error2)

>> install CUDA 10.0

: pip install tf-nightly

 

* error3)

>> pip install easydict

 

- get some .pb files & run the demo script

 

image_demo.py 결과

 

video_demo.py 결과

 

3. Train on my own dataset

3.1 Train VOC dataset

- 디렉토리 구성

 

 

- edit ./core/config.py

 

(2) train from COCO weights(recommend):

* error4)

>> pip install tqdm

 

- 실행

 

- test and evaluate

__C.TEST.WEIGHT_FILE            = "./checkpoint/yolov3_test_loss=44.4873.ckpt-1"

 

>> 수정

 

evaluate.py 결과

>> 해결해야 함.

'인공지능' 카테고리의 다른 글

[PBL] 데이터셋 라벨링  (0) 2019.10.01
[labeling] Yolo_mark로 라벨링  (1) 2019.09.04
Violence Detection 분석  (0) 2019.08.09
Violence Detection 실습  (0) 2019.08.05
Tensorflow Object Detection API with custom training data  (0) 2019.07.05

댓글