Comments (3)
Hi,
that's a pity. It seems the repository does not exist anymore. It was a nice place to get the data. So now, you'll need to gather the data from somewhere else. If you find links and folder structures, I might be able to help you with renaming and arranging
from kiss.
I already found all the datasets, but unfortunately, none of them correspond to NPZ.≥﹏≤
The datasets:
cute80: https://drive.google.com/file/d/1GFZYAnf2_GZzX-_fhtKwhZlM9c8j6Zr2/view?usp=sharing、
SVT-SVTP: https://drive.google.com/file/d/18MrA2yQsTrOsCM826JAISBT_N2GQOc4i/view?usp=sharing
ICDAR2013: https://drive.google.com/file/d/1MbbprTNEbsSfTyXVHtONQFZg470i4wtb/view?usp=sharing
ICDAR2015: https://drive.google.com/file/d/1Ub6a1drjor6oFcqa_q2zFIy3Ad6Mc4TW/view?usp=sharing
IIIT5K: https://drive.google.com/file/d/1LHowumaiKuZujpgWRbmNzPXEQG75wNKc/view?usp=sharing
from kiss.
Alright, so you got the data that is great!
Now, you'll just need to prepare the npz
files. Preparing them is actually quite simple.
You'll need to create 4 numpy arrays:
- an array that you call
num_words
. This array has only one element and is of typeint
. The element should be the max. number of characters per image (it is callednum_words
because the network thinks that each character is a word). In our experiments we always set this to23
. - an array with one element of type type
int
. You call thisnum_chars
. Here the value of the element should be1
because we havenum_words
words of one character during training. - an array called
file_name
of type string. Here, you concatenate the relative path to all image files that you want to use for evaluation. - another array of strings called
text
. This time with the word in each image. Make sure that the indices align. So the word at index 1 in the arraytext
should correspond to the correct image file at index 1 in the arrayfile_name
.
Once you have all of these arrays, you just need to save them (let me show you an example):
# create the arrays
data = {
"num_words": ....,
"num_chars": ....,
"file_name": ....,
"text": ...
}
# now we save everything
with open("destination.npz", 'wb') as f:
numpy.savez_compressed(f, **data)
from kiss.
Related Issues (18)
- MultiGPU HOT 1
- How do you generate the mask in transformer model and process text labels to "class_id" ? HOT 3
- What is the inference speed of KISS ? HOT 4
- Change the num_words_per_image without training again HOT 1
- Some of files in the link are not downloadable. HOT 3
- Use pretrained model and continue training on own data HOT 1
- [Help] Using Pretrained Model HOT 6
- what is gt.mat? HOT 2
- Pretrained model & Dataset HOT 1
- Windows fatal exception: Access violation
- cannot install without a GPU HOT 9
- Training Text-Detection HOT 6
- Is it possible to use KISS on a test image? HOT 11
- SVT Evaluation HOT 1
- link to download SynAdd dataset.? HOT 4
- mjsynth.npz only has first letter of each word in "text" HOT 2
- Loss Functions HOT 25
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from kiss.