search

bhargavaurala / accessmath-icfhr2018

Lecture Video Summarization by Extracting Handwritten Content from Whiteboards
GNU General Public License v3.0
18 stars 4 forks source link

The result of text detection is not good when I only use 320*320 input size. Is the result is right? #7

Open RORO11 opened 3 years ago

RORO11 commented 3 years ago

I have a gpu(12GB), and I run text detection(320 *320) that the gpu is used 9GB. Is it right?

RORO11 commented 3 years ago

FileNotFoundError

FileNotFoundError: [Errno 2] No such file or directory: 'output/classifier/RF_T16_D12_F32_w7x7.dat'. Where can I download it? Thanks.

kdavila commented 3 years ago

FileNotFoundError

FileNotFoundError: [Errno 2] No such file or directory: 'output/classifier/RF_T16_D12_F32_w7x7.dat'. Where can I download it? Thanks.

Hello there,

This file is a Random Forest binarizer trained using the methods originally proposed in the paper: Davila, K., & Zanibbi, R. (2017, November). Whiteboard video summarization via spatio-temporal conflict minimization. In 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR) (Vol. 1, pp. 355-362). IEEE.

The file represents the pickled classifier so it is hard to share a trained version that will run right away on your system. However, you can still train it locally using the keyframe annotations and the following script:

https://github.com/adaniefei/AccessMath_Pose/blob/master/train_ml_binarizer.py

RORO11 commented 3 years ago

@kdavila The advice helps me a lot. Thanks.